Anna K. Geology and geochronology of the middle Behrensmeyer Kipsaramon site complex, Department of Paleobiology, Muruyur Beds, Tugen Hills, Kenya National Museum of Natural History, Smithsonian The Muruyur Beds are a substantial sedimentary deposit within Institution, Washington, DC a middle Miocene sequence of mafic volcanic flows associated 20560-0121, U.S.A. E-mail: with early stages of rifting in the central Kenyan Rift Valley. They [email protected] are best represented in the Muruyur region, near Bartabwa, north of Kipsaramon, where dates range from 16·0 to 13·4 Ma. At Alan L. Deino Kipsaramon, located about 10 km south of Muruyur along the crest Berkeley Geochronology of the Tugen Hills, the upper Muruyur Beds are absent and the lower Center, 2455 Ridge Road, part can be divided into three members. Important fossil sites within Berkeley, California 94709, Member 1 are dated between 15·8 and 15·6 Ma, and within Member U.S.A. E-mail: 3 between 15·6 and 15·4 Ma. BPRP#89, in Member 1, is a bonebed [email protected] at least 2500 m2 in areal extent and up to 30 cm thick, which constitutes one of the richest concentrations of in situ fossil vertebrate Andrew Hill bones in eastern Africa. BPRP#91, at approximately the same level at Department of Anthropology, BPRP#89, is the source of a hominoid talus and other and Yale University, Box 208277, bird fossils. In Member 3, BPRP#122 has produced specimens of at New Haven, Connecticut least five individuals of the hominoid Equatorius, including a partial 06520, U.S.A. E-mail: skeleton. The Muyuyur Beds were deposited near the western margin [email protected] of a lake that was formed during the early stages of faulting and volcanism in the African Rift system. The bonebed in Member 1 appears to represent the influx of fluvially transported vertebrate and John D. Kingston plant remains into a shallow portion of the lake. Elements of the fauna Department of Anthropology, as well as stable isotopes that indicate both forest and more open Emory University, 1557 environments occurred in proximity to the lake during the time of Pierce Drive, Atlanta, Georgia deposition of Member 1. 30322, U.S.A. E-mail: [email protected] Jeffrey J. Saunders Illinois State Museum, RCC 1011, East Ash Street, Springfield, Illinois 62703, U.S.A. E-mail: [email protected] Received 28 February 2001 Revision received 13 June 2001 and accepted 14 June 2001

Keywords: Kipsaramon, Miocene, Kenya, Tugen Hills, stratigraphy, Journal of Human Evolution (2002) 42, 11–38 geochronology, hominoid, doi:10.1006/jhev.2001.0519 Equatorius. Available online at http://www.idealibrary.com on

Introduction in the Baringo District of Kenya. The fossil assemblages are concentrated primarily at There are a number of very rich fossil sites in two horizons within the Muruyur Beds, and the area around Kipsaramon, Tugen Hills, contain a varied vertebrate fauna, including

0047–2484/02/010011+28$35.00/0 12 . .  ET AL.

Figure 1. (a) Overall context of the Kipsaramon site within the East African Rift system, showing other hominoid-bearing sites of middle Miocene age (solid circles). Heavy black lines indicate major faults associated with the rift system. monkeys and hominoid primates such as first described by Chapman, who also first Equatorius africanus (Hill et al., 1986, 1991; recorded fossils from the unit (Chapman, Hill & Ward, 1988; Hill, 1989, 1994, 1995, 1971; Bishop et al., 1971). Exposures of 1999; Brown et al., 1991; Ward et al., 1996, these sediments extend N–S along the sharp 1999; Kelley et al., 2002; Sherwood et al., crest of the Tugen Hills fault block, occur- 2002), as well as plant, ostracod, and mol- ring on the steep terrain of the eastern luscan remains. The Muruyur Beds were flank where the Saimo fault-scarp faces the     13

Figure 1. (b) Mapped extent of the Muruyur Beds (dark gray areas); dashed-line box shows area enlarged in C. (c) Topographic map of the portion of the Tugen Hills indicated in (b), showing locations of the Kipsaramon fossil localities. Dashed-line box shows the area enlarged in Figure 4.

Baringo basin, and on the western flank Kipsaramon area and an interpretation of which slopes more gradually westward the paleoenvironmental setting of the fossil- towards the Kerio Valley [Figure 1(a), (b)]. bearing sedimentary units. We present Fossiliferous localities occur along the preliminary hypotheses concerning the ridge itself and in small patches of exposure genesis of the two main fossil assemblages, on the heavily vegetated, and now but a comprehensive treatment of their intensively farmed, west-facing hillsides taphonomy and paleoecology is beyond the [Figure 1(c)]. scope of this article. We also discuss, but do The Kipsaramon geology and faunal not resolve, several problems concerning remains are important because of the pre- correlations of different outcrops of the cision of the dating, the quality of the verte- Muruyur Beds outside of the Kipsaramon area brate record, and the geographic position of and the stratigraphic position of the unit in this record between the hominoid-producing the regional sequence of volcanic formations. sites of Maboko and Fort Ternan in the Winam Gulf in western Kenya and the Turkana Basin and Samburu areas of Geological overview northern and central Kenya. This paper pro- The fossiliferous deposits of the Kipsaramon vides the first comprehensive geological site complex occur near the base of the long and geochronological framework of the Neogene volcanic-sedimentary sequence 14 . .  ET AL.

Figure 2. General stratigraphic section of the Tugen Hills succession, showing the positions of the sedimentary and volcanic units and hominoid sites. The Kipsaramon localities occur in the Muruyur Beds, which interfinger with phonolite flows near the base of the succession. forming the regional topography of the Rift present in the Tugen Hills, is Group II, also and the Tugen Hills. Correlation and radio- known as the Tugen Hills Group. It is com- metric dating of strata throughout the region posed mainly of flood phonolites, and in the have revealed a relatively continuous record Tugen Hills consists of the following for- of geology and faunas for this time interval, mations: the Sidekh Phonolites Formation extending from over 16 Ma to the present (Chapman, 1971; originally the Saimo (Figure 2). Various aspects of structure and Phonolites Formation of Martyn, 1969), the the large-scale stratigraphy are described in Noroyan Formation, the Tiim Phonolites Bishop et al. (1971), King & Chapman Formation, the sedimentary Ngorora For- (1972), King (1978), Chapman et al. (1978) mation, and the Ewalel Phonolites. and Hill et al. (1986). Neogene formations The Sidekh Phonolites Formation con- in the northern Rift, at least south of about tains a number of sedimentary members, 130N, can be divided into five groups. The largely unfossiliferous. In the north the Group that concerns us here, the lowest formation totals 1200 m in thickness,     15 comprising 10–12 phonolite flows and two lying the Ngorora Formation at Kabasero of sedimentary intercalations. Further south it 12·3 Ma. A re-dating of this sample using is about 700 m thick, and includes 5–6 flows 40Ar/39Ar step-heating resulted in a date of and up to 80 m of interbedded sediment. 13·240·13 Ma (Deino et al., 1990, cor- The Sidekh Phonolites represent some of rected for revised age of the Fish Canyon the earliest volcanics associated with the Tuff standard, Renne et al., 1998). developing rift valley in this part of East Africa. Geology of the Muruyur Beds The lavas of the overlying Tiim Phonolites Formation are about 1050 m The first allusion to the Muruyur Beds thick in the center of the Tugen Hills area, (spelled Muruywr) is in Martyn (1969) who thinning to about 300 m by 1N. There are anticipated Chapman’s later description of between 11 and 15 flows in the type area, the unit (Chapman, 1971) in suggesting that around Tiim mountain about 12 km south these beds are lateral equivalents of sedi- of Kipsaramon. The Formation was divided ments within the Tiim Phonolites near by Chapman (1971) into two members, the Kipcherere in the neighboring area to the lower Atimet Trachyphonolite Member, south. In formally naming the unit, and an upper Kamuiton Phonolites Chapman (1971) gave a detailed account of Member. The Muruyur Beds were described their stratigraphy and lithologies. He by Chapman as a sedimentary intercalation described the beds as ranging in thickness within the Kamuiton Phonolites Member. from less than 10 ft (3 m) near Kamelon to The Sidekh Phonolites Formation is 900 ft (274 m) in the Muruyur section itself probably at least partially equivalent to the [Figure 1(b)], and estimated that 500–600 ft Elgeyo Formation, the Chof Phonolites (152–183 m) of this discrepancy was due to Formation, and the sedimentary Tambach contemporaneous movement along a single Formation of the Elgeyo Escarpment, which fault, leading to differential sedimentation. forms the present western boundary of the Apart from the occurrence of fish fossils Rift in this region (Lippard, 1973; Chapman in various parts of the section, he noted et al., 1978). The Tiim Phonolites For- one fossil site (2/304) that produced a mation is, on the basis of radiometric deter- crocodilian vertebra. The age of the unit was minations (Chapman & Brook, 1978; Hill constrained by virtue of its stratigraphic et al., 1985; Deino et al., 1990), generally placement within the Tiim Phonolite equivalent to the Uasin Gishu Phonolite Formation. Formation of the Elgeyo. The first mention of the unit other than in K/Ar determinations reported in unpublished dissertations was in Bishop Chapman & Brook (1978) give dates for the et al. (1971), where the spelling was changed Sidekh phonolites of 16·4 Ma on a lower to Muruyur. This recorded Chapman’s flow, and 14·8 Ma on an upper flow. [These fauna of fish and crocodile and suggested and all other dates quoted here have been the sediments were probably between 13 corrected for new constants where necessary and 14 Ma in age. Pickford (1975) made according to Ness et al. (1980).] For the some further investigations, discovered Tiim phonolites, Chapman & Brook (1978; additional fossil sites and expanded the also see Bishop et al., 1969; Baker et al., faunal list. Regarding the section at 1971) give 13·3 Ma, 14·7 Ma and 15·3 for Muruyur he remarked: ‘‘measurement of lower flows, and 11·7 Ma, 13·1 Ma and stratigraphic sections . . . is difficult and 13·5 Ma for upper units. Hill et al. (1985) hazardous, accuracy being subject to the give a date for the Tiim phonolite under- daring of the surveyor’’ (Pickford, 1975). 16 . .  ET AL.

However, whether this assertion implies fauna, but are difficult to reconcile with the we should have more confidence in his putative position of the Muruyur Beds in measurements rather than Chapman’s (or the regional stratigraphy, and with dates ours) is unclear. Chapman et al. (1978) dis- from elsewhere on the Tiim phonolites. The cussed the large-scale stratigraphy and struc- dates are in conflict with previous dates on ture of the Tugen Hills range, including the Tiim phonolite flows, and are more in regional equivalence of units. They contin- accord with previous dates on the under- ued to allocate the Muruyur Beds to the lying Sidekh Phonolites Formation. This Tiim Phonolite Formation. This was the poses a problem concerning which volcanic position taken in subsequent reviews that formation properly includes the sedimentary discuss the unit (Hill et al., 1985, 1986; Hill, Muruyur Beds. Later discussions repeat this 1987). ambiguity without resolving it (Hill, 1995, Hill & Ward (1988), in referring to a 1999); nor is it satisfactorily resolved in this hominoid from one of the Kipsaramon sites paper. (KNM-MY 24: BPRP#91), noted Jacob’s observation (personal communication) that Muruyur stratigraphy in the the presence of Diamantomys in the Kipsaramon area Muruyur fauna implied an age for the unit nearer 16 Ma than 13 Ma. Pickford (1988) The stratigraphy, structure, and paleogeo- gave an account of the Muruyur Beds where graphical relationships of the Muruyur Beds he too referred to the significance of at Kipsaramon were reconstructed using Diamantomys. But at the same time, he standard geologic mapping procedures com- located the Muruyur Beds stratigraphically bined with numerous geological trenches between flows two and three, or three and dug in strategic areas of outcrop. Each four, of the Elgeyo escarpment sequence of trench was given a standardized designation, Plateau Phonolites. These phonolites, which e.g., BPRP#89A, with BPRP being the pre- are dated at around 13·5 Ma, are equivalent fix used for all Baringo Paleontological to the Tiim Phonolites and lie above the Research Project sites and the number being Atimet Trachyphonolite dated at 14·3 Ma. the outcrop area (most of which are also In summary, the early regional strati- fossil localities), with the letter suffix desig- graphic work placed the Muruyur Beds in nating the trench itself. Most trenches the Tiim Phonolites Formation, and radio- exposed 3–10 m of strata, and individual metric dates on that formation, rather than beds were documented at a scale of centi- on the Muruyur sediments themselves, led meters to decimeters. The composite to the conclusion that they were between 14 section for Kipsaramon (Figure 3) was con- and 13 Ma in age. structed using marker horizons and distinc- Hill et al. (1991) provide the first dates tive stratigraphic intervals to link together directly relating to Muruyur strata in the the trench sections. Kipsaramon area. Anothoclase phenochrysts The geographical relationships of the extracted from a phonolite flow below the many separate fossiliferous exposures at Kipsaramon bone-bed (in vicinity of Kipsaramon are shown on a large-scale map BPRP#89D) gave a 40Ar/39Ar single crystal constructed by distance, elevation, and com- date of 15·8 Ma, and a determination of pass measurements along paths and roads 15·6 Ma on anothoclase was obtained on an that link discontinuous exposures through ignimbrite immediately above the bonebed areas of bush and farmed plots (Figure 4). (at BPRP#89A) (Table 1). These dates are This map permitted documentation of more in line with suggestions based on the changes in lithology along the roads, Table 1 Analytical data and results of radiometric age determinations

Unit Sample Lab ID no. N1 MSWD2 Ca/K1 S.E.M.340Ar1/39 Ar1 S.E.M.4 J102 1 S.D.5 Age (Ma)1 S.E.M.6

Muruyur type section:       ‘‘Upper’’ Tiim phonolites MUR-4 4026 13/1 1·273 0·0620·038 0·46610·0010 1·5570·006 13·12(0·02)0·06 Muruyur Beds (top) MUR-3 4018 6/1 1·206 0·0380·014 0·47560·0012 1·5540·006 13·37(0·03)0·06 Muruyur Beds (base) MUR-2 4019 12/1 1·260 0·0470·029 0·55360·0010 1·5610·006 15·63(0·02)0·07 ‘‘Lower’’ Tiim Phonolites MUR-1 4025 13/0 0·976 0·0460·102 0·56670·0013 1·5600·006 15·98(0·02)0·07 BPRP#122: Tuff (1·0 m above site) KIPS93-2 7993 15/1 0·709 0·0090·001 2·51200·0020 0·34020·0007 15·45(0·01)0·03 Tuff (3·5 m below site) K122/2 4022 15/0 0·862 0·0320·020 0·54600·0009 1·5550·006 15·35(0·01)0·06 Tuff (8·0 m below site) K122/3 4020 13/4 1·066 0·0210·005 0·54830·0013 1·5580·006 15·45(0·02)0·07 Tuff (10·2 m below site) K122/1 4029 9/1 2·152 0·1260·030 0·54570·0014 1·5620·006 15·41(0·03)0·07 Tuff pinchout (Gorge): ‘‘Lower’’ Tiim phonolites KIPS-1 4730 8/0 1·241 0·0290·009 4·18000·0030 0·20960·0006 15·84(0·01)0·05 Biryokwonin Tuff KIPS-2 4732 8/0 1·053 0·0210·009 4·09030·0029 0·21070·0006 15·58(0·01)0·05 BPRP#89 bonebed: ‘‘Lower’’ Tiim phonolites AH87-1A 1309 12/0 0·979 0·0220·005 1·22170·0014 0·71520·0036 15·80(0·02)0·08 Tuff (1 m below bonebed) KIPS93-3 7995 12/0 1·931 0·0120·007 2·61320·0021 0·33780·0007 15·96(0·01)0·04 Biryokwonin Tuff 89A-1 4736 7/0 0·983 0·0590·031 4·10900·0039 0·21020·0006 15·62(0·01)0·05 Biryokwonin Tuff KSM87-1 1310 10/0 1·473 0·0610·039 1·19980·0020 0·71680·0036 15·55(0·02)0·08

1Number of analyses expressed as total/number rejected. 2Mean sum of weighted deviates. 3Ca/K is calculated from 37 Ar/39 Ar using a multiplier of 1·96. 440Ar* refers to radiogenic argon. 5J is neutron fluence parameter. 6Errors in parentheses do not incorporate error in J; errors outside of parenthesis incorporate error in J after weighted-mean age calculation from individual analyses. Weighting factor is the inverse variance (Taylor, 1982). =5·5431010 y1 . Isotopic interference corrections for Lab IDs 13XX: (36 Ar/ 37 4 6 39 37 4 5 40 39 4 4 Ar)Ca =2·5810 610 ,( Ar/ Ar)Ca =6·710 310 ,( Ar/ Ar)K =8·010 710 ; interference corrections for Lab IDs 40XX: 36 37 4 6 39 37 4 5 40 39 2 2 ( Ar/ Ar)Ca =2·5810 610 ,( Ar/ Ar)Ca =6·710 310 ,( Ar/ Ar)K =2·410 210 ; interference corrections for Lab IDs 36 37 4 6 39 37 4 5 40 39 2 2 473X: ( Ar/ Ar)Ca =2·5810 610 ,( Ar/ Ar)Ca =6·710 310 ,( Ar/ Ar)K =1·110 210 ; interference corrections for Lab 36 37 4 6 39 37 4 5 40 39 2 4 ID nos 799X: ( Ar/ Ar)Ca =2·6410 1·710 ,( Ar/ Ar)Ca =6·7310 3·710 ,( Ar/ Ar)K =1·0810 710 . 17 18 . .  ET AL.

Figure 3. Composite section for the Kipsaramon fossiliferous deposits showing positions of dated samples reported in Table 1.SeeFigure 5 for additional key to lithological codes. Samples with ‘‘Mur’’ prefix are from the section at Muruyur approximately 10 km north of Kipsaramon.

paths, and in the trenches. It also served as a The package of volcanic and sedimentary basis for interpolation between these areas, units dips generally west at 15–25, which is providing an overall schematic view of geo- slightly greater than the angle of the logical contacts and structural features for a topographic slope. This results in older rocks fossiliferous area of roughly 500 m800 m. being exposed higher on the ridge (Figure 5).     19

Figure 4. Site map of Kipsaramon, based primarily on tape-and-compass mapping of localities along the road and paths. Contours are derived from the Bartabwa and Saimo topographic sheets and linked to the tape and compass map using key features from air photographs (scale of 1:24,000). Geological trench localities and/or fossil localities are indicated by black circles; black squares show positions of additional collecting localities for dated rock specimens. The BPRP prefix applies to all sites indicated by the # symbol. K132 is a mollusk locality not presently included in this numbering system.

There are many local variations in the strike observed so far are normal, with down- and dip, some of which appear to represent thrown sides toward either east or west local folding or primary topography within representing extensional displacements as- the sedimentary and volcanic strata, and sociated with the major fault trends that others may be caused by faults obscured by bound the Tugen Hills block. modern soil and vegetation. Numerous well- Lithological features (e.g., grain size, exposed minor faults occur, ranging up from sedimentary structures, color, authigenic less than 1 m displacement. Most of the minerals, provenance of detrital grains and faults are oriented approximately N–S and all pebbles) of the Kipsaramon sediments were 20  .  .  TAL. ET

Figure 5. True-scale (Horizontal:Vertical=1:1) cross sections showing the geological relationships of the major stratigraphic units and the topography between points A–A and B–B in Figure 4. Lithological key given in Figure 3. The BPRP#89 bonebed, represented by the symbol ‘‘’’, is only visible at the points indicated in these cross-sections, but the numerous places where it has been documented throughout the area indicate that it was more laterally extensive than shown here.     21 examined in outcrop, hand-samples, and and buff tuffaceous silts and clays with oc- in a limited number of thin sections. casional sand and mud-clast units (Figure Laboratory study included analysis of grain 3). Following is a general description of the size, CaCO3 content, and clay composition. major units. Much of the local stratigraphic succession consists of fine silt to clay grade clastic material, with lesser amounts of tuffaceous Underlying phonolite and non-tuffaceous sand, occasional pumice The phonolites underlying the sediments are lapilli, reworked clay clasts, and gravel, at least 10 m thick and the uppermost flow which is both volcanic and basement- unit has highly irregular topography on its derived (i.e., quartzite, quartz), in restricted upper surface. The phonolite is deeply lenses. There are sedimentary limestone weathered under the Biryokwonin Tuff on beds and a number of horizons with features the western side of the study area (the Gorge indicating paleosol development (e.g., root section, Figure 4). About 5 m of boulder traces, mottling, slickensides, clay cutans, conglomerate derived from the phonolite lack of bedding, carbonate nodules). The overlies the weathered contact in this sec- paleosol units are typically red to purple red tion, but elsewhere there is remarkably little in color (Munsell 10R4/2), contrasting with coarse clastic material at the contact, and the bright green (Munsell 5GY6/2-7/2, the phonolite forms a smooth surface that is 5Y6/2) of more homogeneous, bedded only moderately weathered. No soil horizon sediment that lacks evidence of subaerial has been observed at the contact other than exposure and pedogenesis. We regard the the weathered phonolite in the Gorge sec- red vs. green color as generally indicating tion. The western end of cross-section A–A original subaerial (Fe-oxidizing) vs. sub- in Figure 5 shows Member 1 pinching out aquaeous (Fe-reducing) depositional set- and Members 2 and 3 in contact with the tings. Red and green silt and clay clasts phonolite. Here, the phonolite dips locally occur together in penecontemporaneous eastward but rises again further east where intraformational conglomerates, indicating it it exposed on the 1920 m contour at the early diagenetic fixing of the color differ- site of AH-87-1. In the northern cross- ence. Thus we use color in conjunction with section (B–B), the phonolite surface is other sedimentological evidence in recon- close to horizontal, when corrected for dip, structing the paleogeography and history of but with many small-scale topographic the Kipsaramon sedimentary deposits. irregularities. The sediments at Kipsaramon represent the lower part of the Muruyur Beds, based on correlations with the Muruyur section. Member 1: fluvial and lacustrine sediments, The Kipsaramon section is about 52 m thick paleosols overall (Figure 3), and can be conveniently Sediments that immediately cover the pho- divided into three members. Member 1 con- nolite generally lack even sand-grade clastic sists of up to 20 m of green, red, and purple debris from the phonolite in exposures clays, silts, and occasional sands and lime- examined so far, except for occasional large, stones, which overlie the basal phonolite. weathered blocks within the upper green Member 2 is a distinctive 1–10+ m thick clays at site BPRP#89F, which are associ- ash-flow tuff referred to as the Biryokwonin ated with a high point on the original surface Tuff, which A.D. named for a school near of the phonolite. This indicates that the high outcrops of the Tuff in the Muruyur type point was exposed subaerially (e.g., as an area. Member 3 consists of 25+ m of green island) while the rest of the flow surface was 22 . .  ET AL. under water. The sedimentary sequence include quartz and feldspars indicating above the phonolite varies considerably, but fluvial input from igneous basement rocks as can be characterized overall as consisting of well as volcanic sources. a lower sequence of 5–10+m of red, purple, Analysis of the clays reveals that the and green clays with some pedogenic modi- primary constituent in both the red and fication near the base and chert nodules the green colored clays is illite. Kaolinite toward the top, followed by 5–7 m of green and chlorite are absent, and smectite occurs clays and sandy clays with thin beds of tuff, only in small amounts in the green clays. limonitic nodules, and occasional coarse Smectite is known to be a common units including the BPRP#89 bonebed weathering product of volcaniclastic sedi- (Figures 3 and 6). The sequence of litholo- ments (Retallack, 1991) and would be gies is best exposed at BPRP#89G, where a expected based on other evidence of vol- discontinuous white tuff and pedogenically canic input to the Muruyur deposits. Given modified red clays overlie phonolite, fol- the considerable thickness of the overlying lowed by laminated red and green clay, section, however, it is possible that the ostracod-bearing limestones, variegated illite represents low-grade metamorphic clays (red, purple, green) with paleosols, alteration of original smectitic clays. green silty clays and sandy clay-clast con- Alternatively, the illite could represent pene- glomerates (the most prominent of which is contemporaneous alteration of smectite in the BPRP#89 bone bed), finally capped by evaporitic lake deposits (Singer & Stoffers, the Biryokwonin Tuff. The micritic lime- 1980). stones include oolites and abundant ostra- The upper part of Member 1 consists of cods, and the chert nodules resemble those green clay with coarser beds of tuffaceous formed in alkaline lakes (Surdam & Eugster, silts, intraformational clay-clast and tuff- 1976). Bedding in these sediments is domi- clast conglomerates, and sands with vol- nantly horizontal and well-preserved on mm canic rock fragments indicating proximity to cm scales; ripple lamination is rare. There to a source of current-derived clastic and is some fine-scale alternation of red and volcaniclastic materials. The BPRP#89 green coloration that conforms to the bonebed and associated leaf bed (Hill lamination, but most beds are dominantly et al., 1991) occur in a widespread set red or green. Bioturbation in the laminated of exposures of this coarser sediment, clays is minimal, suggesting relatively rapid which is between 0·3 m and 1·2 m in deposition or a lake chemistry unfavorable total thickness. Immediately below the to benthic fauna. Bright red clays and bone-bearing unit, there is evidence for well-developed paleosols are more com- bioturbation in shallow subaquaeous mon in the north end of the exposures conditions (indicated by burrowing in (BPRP#89G, BPRP#91), where Member 1 green clays at the main excavation area, is also relatively thin, suggesting that more BPRP#89A) as well as pedogenesis on subaerial conditions were associated with a subaerial land-surface (indicated by a locally higher original elevations of the red paleosol with CaCO3 nodules at underlying phonolite. Greater thicknesses of BPRP#89G, 400 m to the north of Member 1 green and purple clays with less BPRP#89A). Red clays below the evidence of pedogenic modification occur in Biryokwonin Tuff 180 m WSW of the central and southern part of the area. BPRP#89A also suggest more subaerial Sand-grade material is a relatively minor conditions associated with a topographic constituent except in the BPRP#89 bone- high on the underlying phonolite just prior bearing unit, but the clastic sediments to the time of bone-bed formation.     23

Member 2: the Biryokwonin Tuff exposure and weak paleosol development. The tuff is a distinctive, yellow-gray, There is no evidence of the red or purple anorthoclase-bearing, pumiceous, vitric, coloration that characterizes paleosols in poorly welded ignimbrite, less than 1·0 m Member 1. One prominent bed of black thick in the southern part of the Kipsaramon chert exposed in the road near BPRP#89F exposures, increasing to over 10 m in the could represent diagenetic alteration of the north and pinching out to the west above silica-rich volcanic sediments, but might the phonolites in the Gorge section. The also indicate primary deposition of chert in a equivalent unit in the Muruyur section local high-alkalinity playa or pond. The tuffs 10 km to the north is 20 m thick and more include small pumice fragments and feld- welded than at Kipsaramon, suggesting a spars and show variable degrees of dia- source nearer Muruyur. At Kipsaramon, genetic alteration. XRD analysis of one of the upper 3–5 m of sediments below the these tuffs showed alteration to analcime, Biryokwonin Tuff are generally green and which also indicates an alkaline lake or early fine-grained with thin beds of altered tuff diagenetic conditions (Hay, 1976). Many throughout the area, indicating lacustrine appear to be primary air-fall tuffs, with conditions prior to deposition of the ignim- unsorted or vertically graded textures, brite (Figures 3 and 6). The ignimbrite itself while others have a well-sorted sand-grade appears to have flowed into this lake; there component indicating reworking. are no indications of emergence (e.g., mud cracks, clay breccia, root marks) immedi- ately prior to deposition, and the lower Geochronology surface of the Tuffs is irregular, with large downward bulges suggesting deformation We have applied the single-crystal, 40Ar/ into underlying unconsolidated sediment. 39Ar laser-fusion dating technique to 14 samples of tuff and phonolitic lava taken Member 3: clays and volcaniclastics both from the area of the Kipsaramon site complex and also from the section at The highly irregular upper surface of the Muruyur itself. These new geochronological Biryokwonin Tuff formed numerous areas data permit us to document the age of the where local reworking resulted in finely Muruyur Beds and critical paleontological laminated tuffaceous pond deposits. Gastro- sites directly. pods and bivalves are preserved as silicified Anorthoclase grains were hand-picked for steinkerns, generally within the first meter of 40Ar/39Ar analysis from crushed pumice sediments above the upper surface of the wherever possible, or from whole-rock tuff. Otherwise, based on exposures of samples if pumices were rare or unavailable. Member 5 below and including locality The feldspar concentrates were washed in BPRP#122, the sediments overlying the 5% HF and 10% HCl solutions in an ultra- Biryokwonin Tuff consist of green clays with sonic cleaner for 5 min each. After examin- limonitic silts, occasional sandy or silty clay ation of the feldspar populations under a beds, and finely laminated to massive beds binocular microscope for clarity and free- of fine- to coarse-grained altered tuffs. dom from inclusions, the best grains were Limonitic nodules are less common than in loaded into aluminum sample holders and Member 1. Some of the green clay beds irradiated for 3–22 h in the core of a nuclear have reworked tuff clasts, root traces, reactor (two reactors were used; see below). burrows, and other evidence of bioturbation After a delay of up to several months for indicating shallow aquatic or subaerial radiological cooling, the individual grains 24 . .  ET AL.     25 were fused under ultra-high vacuum with a The replicate single-grain analyses for indi- tightly focused continuous laser beam (Nd- vidual samples typically generate simple YAG or Ar-Ion), followed by measurement gaussian-like distributions, as illustrated by of five Ar isotopes in a low-blank, high- age-probability distributions (Figure 7). In sensitivity noble gas mass spectrometer one case (K122/3), xenocrystic material can (MAP 215 or 215-50). From six to 15 be recognized as a minor population of phenocrysts from a given sample were ana- outliers that are distinctly older than the lyzed to test reproducibility and sample primary, young population of grains; these homogeneity. were omitted from the weighted-mean age. Two different monitor minerals (age In addition, those few analyses that fell more standards) were co-irradiated along with the than two standard deviations beyond the unknowns during the course of the project; weighted mean age of a sample population both were sanidine of mid-Tertiary age. The were also omitted to avoid skewing mean neutron fluences in samples with laboratory ages, though these grains are probably not IDs 1309 and 1310 (Table 1) were moni- xenocrysts. These outliers may reflect alter- tored with sanidine from the Fish Canyon ation or the presence of inclusions, or simply Tuff of Colorado, using a reference age of analytical variability. 28·02 Ma (Renne et al., 1998). All other samples were co-irradiated with an internal Section at Muruyur sanidine standard, WHRN-1J from the The base of the sedimentary and volcani- Pahranagat Tuff, Nevada. This material has clastic section at Muruyur, about 10 km an age of 22·7820·007 Ma relative to the north of Kipsaramon [Figure 1(b)], near Fish Canyon Tuff at 28·02 Ma (Best et al., Bartabwa, rests on phonolitic lavas and 1995, revised for Renne et al., 1998). In all, agglomerates that have been attributed four separate irradiations were performed, to the Tiim Phonolites Formation. one in the CLICIT Cd-lined in-core facility Anorthoclase phenocrysts from the upper- of the Oregon State University TRIGA most lava flow beneath the section at reactor and three in the hydraulic rabbit Muruyur produced a single crystal total in-core positron of the Los Alamos Omega fusion (SCTF) 40Ar/39Ar age of West reactor. In each case the vertical 15·980·07 Ma (Table 1, Sample MUR- and lateral neutron flux gradients were 1). This date conflicts with previous age accounted for by measurement of multiple determinations on lower Tiim Phonolite sanidine monitor positions in the irradiation flows in the type area at Tiim mountain, package. For additional details, see the 12 km south of Kipsaramon, and is closer to discussions of the single-crystal 40Ar/39Ar previous dates on the underlying Sidekh analytical procedure available in Deino & Phonolites Formation. However, the date of Potts (1990) and Deino et al. (1990). 15·980·07 Ma is in agreement with our Analytical results of the dating experiments earlier 40Ar/39Ar dates on the underlying are summarized in Table 1. volcanics at Kipsaramon (Hill et al., 1991). In general, interpretation of the results of As will be shown below, the ‘‘Tiim Phono- the dating experiments is straightforward. lites’’ below the Muruyur Beds in the Gorge

Figure 6. Fence diagram of the Kipsaramon deposits, showing lateral changes in the context of the BPRP#89 bonebed within Member 1 and changes in thickness of the Biryokwonin Tuff (Member 2). The phonolite underlies Member 1 throughout the area but is not documented except where shown on the measured sections. Lithological details are shown for sections documented using geological trenches. Bone symbols in parentheses indicate surface rather than in situ occurrences. 26 . .  ET AL.     27 section at Kipsaramon are about the same believed to be equivalent to the Biryokwonin age as other determinations in this area, Tuff (Member 2; Figures 3, 5 and 6), but suggesting that the entire ‘‘lower’’ Tiim thicker and more welded as a result of Phonolites Formation may be older than accumulation in a local basin or closer about 15·8 Ma. The cause of this discrep- proximity to the eruptive source. In the ancy between the new 40Ar/39Ar date and excavation area at Kipsaramon, up to 20 m previous younger conventional ages on of sediments including the BPRP#89 ‘‘lower’’ Tiim flows is not known; it may bonebed lie between the phonolites and the reflect the inherent improved precision Biryokwonin Tuff (Member 1 in Figure 3). and geologic accuracy of the anorthoclase These predominantly fine-grained lacustrine single-crystal 40Ar/39Ar method, as com- and subaerial deposits are not represented in pared to conventional whole-rock K-Ar the lower Muruyur Beds at Muruyur. dates on lavas. On the other hand, it may be The thick sedimentary and volcaniclastic that substantially different age lava flows are sequence of the Muruyur Beds above the being sampled. If so this may imply errors in ignimbrite at Muruyur is capped by a the regional stratigraphic interpretation; biotitic phonolite lava (MUR-4) attributed these dates suggest that the ‘‘lower Tiim’’ to the upper Tiim Phonolites Formation. flows at Kipsaramon and at Muruyur may in This lava is dated at 13·120·06 ma, an fact be the Sidekh Phonolites. age consistent with previous dates on upper The 15·980·07 Ma lava, which is the parts of the Tiim Formation including our uppermost lava flow beneath the Muruyur own 40Ar/39Ar ages obtained on phonolites section, is overlain by a covered interval of that just underlie the Ngorora Formation about 25 m, followed by approximately approximately 17 km away (Deino et al., 20 m of phonolitic agglomerate. This is 1990). A sample of medium- to fine-grained overlain, in turn, by a 20 m thick, bedded tuff 4 m below the phonolite near white to green, pumiceous, anorthoclase- the top of the Muruyur Beds also gave a bearing, vitric welded ignimbrite (MUR-2, concordant age of 13·370·06 Ma (Sample Table 1), considered to represent the MUR-3). base of the Muruyur Beds in this section. Thus, in the section at Muruyur, the sedi- The age obtained for the ignimbrite, mentary and volcaniclastic deposits span an 15·630·07 Ma, is consistent chronostrati- interval of about 2·5 Ma, whereas, as will be graphically with the underlying dated phono- shown below, in the vicinity of the lite and is close to ages obtained for the Kipsaramon site most of the younger part of Biryokwonin Tuff in the excavation area, as the Muruyur section either is missing and described below. The MUR-2 ignimbrite is presumed eroded away prior to emplacement

Figure 7. Age-probability spectra and auxiliary plots of 40Ar/39Ar analytical data for all dated samples. The age probabilities are calculated assuming a unit gaussian error for each analysis, followed by summation across all samples of probabilities within narrow age intervals. The dashed line in the age-probability plots represents the case where all analyses are included, whereas the solid line results from elimination of analyses falling more than two standard deviations beyond the weighted mean. The omitted analyses are identified by the open circles in the auxiliary plots above the probability diagrams. The mode of each spectrum is shown near the peak of the curves. Error bars near the bottom represent the standard error of the weighted mean, both with (outer ticks) and without (inner ticks) error in J, the neutron fluence parameter. Values to the right of the error bars represent the standard error of the weighted mean of the included analyses with one-sigma standard error of the weighted mean, incorporating error in J. The auxiliary plots of associated 40Ar/39Ar data include (1) moles 39Ar released during fusion, (2) percent radiogenic 40Ar of total argon (40Ar*), (3) Ca/K ratio, and (4) display of chronologically sorted individual ages with one standard deviation analytical uncertainty. 28 . .  ET AL. of the capping Tiim phonolites or was never Member 2. Three mutually indistinguishable deposited locally. anorthoclase phenocryst SCTF ages were obtained from the Biryokwonin Tuff at Kipsaramon site complex two localities in the study area (Table 1), producing a weighted-mean age of 15·59 Phonolite. Anorthoclase phenocrysts from 0·03 Ma. This age is indistinguishable from two samples from the uppermost exposed the age of 15·630·07 Ma obtained from phonolite flow beneath the sedimentary sec- the ignimbrite at the base of the Muruyur tion at the Gorge (KIPS-1) and at a site Beds at Muruyur, indicating that these two 100 m south of BPRP#89D (AH87-1A) outcrops likely represent the same or closely have been dated by the SCTF method, related flow events. yielding mean ages of 15·840·05 and 15·800·08 Ma, respectively. Although Member 3. The age of Member 3 is con- these ages are indistinguishable, it is not strained by the age on the underlying known whether these outcrops belong to the Biryokwonin Tuff of 15·590·03 Ma, and same phonolite flow unit. The combined by dated units within the member itself. weighted-mean age of these determinations Anorthoclase phenocrysts obtained from is 15·830·04 Ma. This result is indis- four tuff beds in Member 3 were dated by tinguishable at the 95% confidence level the SCTF method, producing, from strati- from the age obtained for the uppermost graphically lowest to highest, 15·41 phonolite flow below the section at Muruyur 0·07 Ma (K122/1; 0·5 m thick buff of 15·980·07 Ma. massive tuff), 15·450·07 Ma (K122/3; white, 2 cm thick massive tuff), Member 1. The age of Member 1 is con- 15·350·06 Ma (K122/2; diffuse 5 cm tuf- strained by the mean 40Ar/39Ar age on the faceous zone with rare biotite), and disconformably underlying phonolite of 15·450·03 Ma (KIPS93-2; altered green 15·830·04 Ma, and a weighted mean age finer- to coarse-grained tuff 10–15 cm on the overlying Biryokwonin Tuff (see thick). These ages are all statistically indis- below) of 15·590·03 Ma, an interval of tinguishable at the 95% confidence level. A 24070 Ka. A 10–20 cm thick, altered Monte-Carlo estimation procedure, using as green massive tuff about 1 m below the a success criterion correct chronostrati- BPRP#89 bonebed, in the middle level of graphic ordering of the three dated units, Member 1, produced an SCTF age on yielded a 95% confidence range for the anorthoclase of 15·960·04 Ma (KIPS93- sequence of 15·58 Ma at the base to 3). This age is statistically older than that 15·36 Ma at the top (K. Ludwig, personal obtained on the underlying phonolite. communication). Examination of the age-probability density spectrum for this sample (Figure 7) may The fossil localities provide an explanation; the sample demon- strates subdued multimodality in ages, and The Kipsaramon site complex is character- may consist largely of older, contaminant ized by a diverse and well-preserved array of xenocrysts. Given this uncertainty, the best fossil materials. The most important sites for maximum age for the bonebed is defined by in situ vertebrate remains are the two locali- the age of the underlying phonolite at ties BPRP#89 and BPRP#122 (Figures 3 15·830·04 Ma, and the best minimum and 6), at two distinct stratigraphic levels. age at 15·590·03 on the overlying There also are a number of other in situ and ignimbrite. surface occurrences of and other     29 fauna, some associated with the BPRP#89 Table 2 Faunal list for the Kipsaramon bonebed level but others representing lower or higher (BPRP#89), compiled by AH and Rebecca Fisher and based mainly on material from the excavation source beds within Members 1 and 3. Fish at BPRP#89A remains, consisting of small bone fragments and occasional scales, are moderately com- Reptilia Crocodilidae mon in the green clays and silts of Member gen. et sp. indet. (31021, 31022) 1 and have been documented in Member 3 Trionychidae gen. et. sp. indet. (31017) as well. Fossil plant remains consisting of ?Pelomedusidae poorly preserved leaves and stems are gen. et sp. indet. (31027) abundant in silts immediately above the Mammalia BPRP#89 bonebed at BPRP#89A and D, Primates Victoriapithecidae but have not been documented elsewhere in Victoriapithecus macinessi (31013, 31014, 31015) the area. Small fragments of silicified wood Family undet. cf. Proconsul major (18690) also occur in Member 1 silts and sandy Kalepithecus (18689) clays. Casts of gastropods and occasional Proboscidea Deinotheriidae bivalves occur in the lower part of Member gen. et sp. indet. 3, particularly at geological sampling site (presumably Prodeinotherium hobleyi) (31011) Amybeledontidae K132 near BPRP#89A, immediately above Protanancus macinnesi the Biryokwonin Tuff (Table 2). Ostracods Choerolophodontidae Choerolophodon kisumensis are abundant in the limestone beds of lower Perrissodactyla Member 1 (at BPRP#89G). Coprolites, Rhinocerotidae probably from fish and crococile, are com- gen. et sp. indet. (31001, 31002, 31005) Artiodactyla mon through Member 1 and particularly at Hippopotamidae BPRP#91 and BPRP#89F. Kenyapotamus cf ternani (31008) Sanitheriidae Microstratigraphic analysis of Kipsaramon Diamantohyus africanus (30974, 31009, 31026) has concentrated on Member 1, which offers Nguruwe sp. (31010) the best exposures within the area (Figure Giraffidae 6). A single measured section through Palaeotragus primaevus (31025) Tragulidae Member 3 at BPRP#122 serves as the basis Dorcatherium pigotti (31012) for interpretations of depositional history as Rodentia Sciuridae well as geochronology for this fossil locality. gen. et sp. indet (22387) We have not examined lateral variations in Pedetidae lithology within Member 2, the Biryokwonin gen. et sp. indet. (small) (19437, 19438) Anomaluridae Tuff, except to document changes in thick- Anomalurus parvus (19439: holotype and only ness throughout the area. specimen) Thryonomyidae gen. et sp. indet. 1 (large) (19426, 19427, 19425) BPRP#89 gen. et sp. indet. 2 (small) (22388) Diamantomyidae Controlled excavations that documented Diamantomys luederitzi (MY 99) sedimentological features, as well as bone Myophiomyidae Elmerimys woodi (19429) positions and orientations, were conducted cf. Elmerimys sp. (22394) at BPRP#89A–D in 1987. The density of Cricetodontinidae fossil material at BPRP#89A proved to be so Notocricetodon sp. indet. (22389) great, however, that subsequent excavation Pascal Tassy and JS identified the Amybelodontid from 1990 onwards involved the removal of and Choerolophodontid. Rodent identifications are large blocks for preparation and taphonomic from Winkler (1990, 1992, 2002). Numbers are the accession numbers of specimens used to diagnose documentation in the laboratory. Approxi- that taxon at the site. They should be prefaced by mately 1000 specimens have been recovered ‘‘KNM-TH’’ except where otherwise indicated. 30 . .  ET AL. from both field and laboratory operations so general flow direction within the bonebed far. This work is still in progress, and a unit, it does provide evidence for fluvial detailed report of the fauna and taphonomy depositional processes. All contacts between of the bonebed will be the subject of a future the coarser deposits and the underlying and publication. Here we present descriptions of overlying finer-grained beds are distinct. the sedimentary features associated with the The green clays below the coarse zone are bonebed, an interpretation of its paleo- laminated, homogeneous in texture and environmental context, and a working include minor disrupted bedding, isolated hypothesis concerning its origin. burrow traces and fragmentary fish bones. At BPRP#89A, the 0·3 m thick bonebed These features indicate deposition in a unit is associated with about 1·4 m of silty poorly oxygenated, probably shallow, sub- tuffaceous beds, sandy mud and tuff clast aquatic environment. There are no signs of conglomerates (with clasts up to 4 cm subaerial exposure in the clays, such as diameter, including occasional quartzite desiccation cracks or pedogenic carbonate. pebbles), and sandy silts that form a coarser Thus, the depositional context of the deposit within the homogeneous green and bonebed at the four closely spaced sites in gray clays that characterize the top of the excavation area is similar overall, but Member 1 below the Biryokwonin Tuff variable in details that indicate fluvial input (Figure 6). The 0·2 m thick plant-bearing of coarse materials into a shallow lacustrine unit immediately above the bonebed con- depositional environment. sists of tuffaceous silts and clayey sands, To the north, the interval of silty and with ripple and wavy lamination and com- sandy clays that includes the bonebed is mon burrow traces. At BPRP#89B, 30 m to present at BPRP#89F and BPRP#89G the northwest (Figures 4 and 6), the 0·5 m (Figures 4 and 6), where bones have thick bone-bearing unit is bounded above been collected in situ. In the other docu- and below by green clays, with laminated mented sections (BPRP#89J+K, and tuffaceous silts occurring 0·4 m lower in the BPRP#91C+D), there is no obvious coarse section. The excavation at BPRP#89D, interval within the upper green clays of 47 m south of BPRP#89A, revealed a Member 1, although surface bones that complex of low angle cross-stratified bone occur up to the level shown in Figure 6 and plant-bearing, interbedded sandy silts suggest the presence of a fossiliferous zone and silty clays, which are approximately 0·2– at the equivalent stratigraphic position. To 0·3 m thick overall and bounded above and the south, at BPRP#89E, a 0·4 m thick below by green clays and silty clays. A interval of silts with small-scale mudcracks is narrow, clay-filled paleochannel oriented present within the upper Member 1 clays approximately north–south occurs at the top and occurs at level of the highest occurrence of the fossiliferous deposit. At BPRP#89C, of surface bones. It thus seems likely that about 25 m southwest of BPRP#89D, the this is the equivalent of the bonebed bone-bearing unit lies within green clays, is level. only about 0·1–0·15 m thick and is overlain At BPRP#89F, bones occur in a lami- by 0·05 m of plant bearing silts. A coarser nated, gritty, silty clay with tuff clasts that is silt and mudclast conglomerate 0·3 m thick generally finer than the fossil-bearing occurs about 0·6 m above the bonebed lithology at BPRP#89A–D. Sediments just beneath the Biryokwonin Tuff at below the bonebed level are laminated and BPRP#89C. The only visible cross- there is no evidence for subaerial exposure. stratification occurs at BPRP#89D, and Near BPRP#89F, the bonebed can be although this is too localized to indicate a traced across the road to BPRP#89I as     31 a continuous fossiliferous unit. At initial bone accumulation is unclear, but it BPRP#89G, about 180 m northeast of must have involved large number of BPRP#89F, the bone-bearing unit is a of all sizes. The variable state of the bones 0·70 m thick, poorly sorted clayey sand with themselves, many of which are abraded and abundant mud and tuff clasts. It overlies a weathered, but some of which are fresh, as thin green clay, which in turn rests upon well as the diversity of species, suggest an a red, pedogenically modified clay with attritional rather than a catastrophic death

CaCO3 nodules (Figure 6). Thus, between assemblage. Long-term accumulation of these two sites there is a transition between bones within a fluvial system, followed by subaerial and subaquatic depositional en- unusual flooding that flushed a large vironments (i.e., a lake edge) just below the number of bones out into the lake, would fit bonebed level. This documents the context the known facts. of the bonebed within the initial deposits of Further work on the taphonomic features a local lake transgression represented by the of the bones themselves, body part and laminated green clays and coarser units of taxonomic representation, and spatial pat- upper Member 1. terns within the excavated sites is needed in An estimate for the maximum lateral order to test this hypothesis for the genesis extent of the bonebed, based on both in situ of the bonebed. The interpretation of this and surface documentation, and assuming unusual deposit by Pickford (1988) was as that it represents a more or less continuous follows: ‘‘the extremely rich ‘bone bed’ sheet of fossiliferous material, is on the order appears to have accumulated subaerially as a of 0·25–0·28 km2. A more conservative esti- vast collection of skeletons, principally of mate, based on only in situ documentation at proboscideans, in an area about 150 m BPRP#89A–D, F, I, and G, would be about across and up to 30 cm thick, presumably a 0·09 km2. There were up to 30 bones/m2 at dried-up water hole.’’ We have found no BPRP#89A in the excavation; assuming that evidence for deposition in a subaerial water- the average is well below this at 1 bone/m2, hole context; instead there is clear evidence we calculate that there could be a total of of transport of large numbers of bone to the between 9000 and 28,000 bones (whole and site of final deposition. The documented fragmentary) in the documented extent of lateral extent of the bonebed also is con- the bonebed. siderably greater than Pickford’s estimate. Our interpretation of the bonebed is that it is part of a fluvial influx into a shallow BPRP#91 lake. The stratigraphic and lithologic This locality was one of the original sources features indicate moderate flow velocities in of fossils in the Kipsaramon site complex, an aggrading (not downcutting) fluvial con- including numerous mammal remains, text, and the lateral facies changes suggest a coprolites, and a hominoid talus (Hill & dominant north–south axis to the bone- Ward, 1988; Pickford, 1988; Hill, 1991, bearing flow. The singular nature of the 1994). Outcrops of lacustrine limestones bonebed itself, with only one coarse, bone- and red and green siltstones at BPRP#91 bearing unit topped by plant-bearing silts are exposed along a shallow drainage, and occurring within much finer sediments, indi- fossils occur on the surface of these expo- cates a unique depositional event. The sures (Figures 3 and 4). There is no visible working hypothesis is that this represents the outcrop of the BPRP#89 bonebed, and flushing of an upstream bone accumulation, detailed stratigraphic correlation to trench perhaps via deltaic distributaries, into a sections to the southwest is difficult because shallow, subaquatic setting. The cause of the of lateral facies changes and discontinuous 32 . .  ET AL. exposures. However, overall similarity of the the in situ material, is the most likely source transition from the red, subaerial lower of the surface fossil material. At least five strata to the green, laminated, sub-aqueous individuals of the hominoid Equatorius upper deposits at BPRP#91 to nearby africanus (see discussion below) are repre- BPRP#89J and BPRP#89G, where the sented by the surface and in situ materials. sediments are capped by the Biryokwonin Other surface fauna collected at the site is Tuff, provides the basis for linking the two sparse, consisting only of fish (including gar sequences. Projections using cross-section scales) and fragments of turtle, crocodile, B–B (Figures 4 and 5) indicate that the and proboscidean rib. bonebed should occur near the upper part of The clays associated with the fossiliferous sectoin BPRP#91C+D, at about the level of zone of Member 3 are heavily bioturbated, a thin tuffaceous silt within green laminated with little evidence of original bedding, and clays. Most of the surface bones occur near include reworked tuff fragments. The tuffs or below this level stratigraphically and are generally white, thin, homogeneous topographically. Given the overall scarcity of units with fresh feldspars and appear to be fossil bones in this area other than those airfall in origin. Other than the bioturbated derived from the BPRP#89 bonebed, and clays, there is no evidence for pedogenic the proximity of sites with exposures of the processes. Overall it appears that the fossil- bonebed, previously eroded portions of it bearing unit is part of a sequence of low- are a possible source of surface fossils dis- energy subaquatic deposits in a shallow covered at BPRP#91. However, some fossils lake or swamp where clay deposition was maybe derived from the subaerial to mar- interrupted occasionally by airfall events. ginal lacustrine deposits below this level as Additional taphonomic study is needed to well. The exact provenience of fossils at this determine the processes responsible for the locality cannot be proven without discovery bone concentration. of in situ bones or trace element analysis that links surface bones to a specific source Fauna and environments deposit (e.g., Plummer et al., 1994). A more detailed analysis of the Muruyur BPRP#122 fauna will appear elsewhere, but some Isolated teeth of hominoids were found on a general comments are made here. The fossil relatively flat area at the base of a steep bank fauna from the Kipsaramon site complex is of exposures of upper Member 3. This bank best known from the excavated locality includes a vertical zone of slumped and BPRP#89A and its lateral correlates. disrupted sediments that may be a clastic Conventional exploratory excavations at dike, but strata on either side represent the BPRP#89A started in 1987 and continued same sequence of beds, so there appears to for several field seasons, but this approach be no structural displacement of the other proved unsatisfactory because in the richest units. Subsequently a hominoid mandible part of the bonebed, the sediment is com- and partial skeleton were discovered in situ, posed almost entirely of interlocked clasts of about 7 m higher than the surface accumu- bone of various sizes. For example, a pro- lation. This specimen comes from a clay/tuff boscidean tooth might have in the matrix couplet within 1·8 m of interbedded green between its cusps a tooth of a carnivore, clays and thin white tuffs(Figures 3 and 6). which in turn might have a rodent tooth Careful investigation of outcrop surfaces within the matrix surrounding it. Conse- and geological trenching demonstrate that quently, starting in 1990, we removed this zone, and the unit within it that bears large blocks of bonebed strengthened with     33 hardeners and plaster jacketing, by cracking Table 3 Faunal list of Mollusca collected from them along planes of weakness with jacks. the top of Member 2 downslope from BPRP#89B These blocks were transported to the Mollusca National Museums of Kenya in Nairobi Gastropoda where individual bones were dissected from Ampullariidae Pila ovata the blocks under a microscope by Boniface Melanoides tuberculata Kimeu. In 1987, we recovered the teeth of ?Bartoa nilotica three different species of hominoids in two Pelycipoda Mutelidae days of conventional excavating the surface Pleiodon sp. indet. few centimeters of a single square meter. However, no other hominoid remains have Identifications were provided by Peter Kat, National yet been found among the thousands of Museums of Kenya. bones removed from blocks from this site in the following years. A preliminary faunal list BPRP#122, about 200 ka younger than the for BPRP#89, based largely on material BPRP#89 level. Between 1990 and 1993, from this excavation, shows a fauna that is survey and screening of this locality resulted very diverse taxonomically, and also in terms in a surface collection of 38 isolated homi- of specimen size, ranging from proboscidean noid teeth from a relatively flat area at the palates to rodent teeth (Table 2). base of a steep bank of exposures. In 1993, A fauna including mammals, bird remains Boniface Kimeu discovered a mandible of and many coprolites has been recovered the hominoid in situ higher up on these from locality BPRP#91, which is either at exposures. Excavation revealed that in the same level as the BPRP#89 bonebed or addition to the mandible, much of the upper within the same few meters of Member 1, postcranial skeleton of this individual was just below the Biryokwonin Tuff. This local- also present in the sediment. After collection ity also produced the first hominoid known and preparation it was clear that the partial from the site complex, an astragalus of a skeleton and the isolated teeth from species comparable in size and preserved BPRP#122, belonging to at least five indi- morphology to Proconsul major, although it viduals, were representatives of what had has features in common with an Afropithecus been referred to as ‘‘Kenyapithecus africanus’’ specimen from Kalodirr (Hill & Ward, (Brown et al., 1991). Analysis of the new 1988; Pickford, 1988; Hill, 1991, 1994). material also showed that ‘‘K. africanus’’ had Additional taxa are known from other to be generically distinct from the type levels. Mollusc steinkerns and molds species, K. wickeri. Accordingly a new were collected from the surface of the genus, Equatorius, was created to accommo- Biryokwonin Tuff just downslope from date this occurrence, and also other fossils, BPRP#89B, and are identified in Table 3. such as those from Maboko, that had All are aquatic except for Bartoa, which is a previously been included in Kenyapithecus land snail. P. Kat, who identified them, africanus (Ward et al., 1996, 1999; see also remarks that the presence of Melanoides Brown et al., 1991; Kelley et al., 2000, 2002; and Pleiodon indicates a lake environment Sherwood et al., 2002). Other than an iso- at this time, at the base of Member 3, lated tooth of Nyanzapithecus, and micro- rather than a swamp or marsh (Kat, mammals from sieving, there is little personal communication). additional fossil material from this site. Another taxon so far known only from Much of the paleontological focus at higher in the sequence is Equatorius Kipsaramon so far has been on the homi- africanus, from upper Member 3 at noids (Hill & Ward, 1988; Hill, 1989, 1994; 34 . .  ET AL.

Brown et al., 1991; Ward et al., 1996, 1999; and generally tolerate higher temperatures Kelley et al., 2002; Sherwood et al., 2002) and solar irradiance, drier conditions, and the micro-mammals (Winkler, 1990, greater seasonality, and lower atmospheric

1992, 2002), with other Kipsaramon fauna pCO2 levels (O’Leary, 1988; Farquhar et al., yet to be studied in detail; P. Tassy and JS 1982) than C3 vegetation (primarily trees, are currently working on the proboscideans, shrubs, herbs and temperate or high altitude which are abundant in the BPRP#89 grasses). C4 plants are associated with bonebed. breaks in closed canopy low-latitude forests When compared to other Miocene sites in or woodlands or more open woodland as East Africa the Kipsaramon fauna is very well as grassland habitats. Isotopic analyses similar to that from sites on Maboko, which of pedogenic carbonate collected from is about the same age. It lacks many of the Member 1 red to purple red paleosol elements shown in slightly later sites such as horizons in subaerial portions of the section Ford Ternan (see, for example, Harrison, at and adjacent to BPRP#91 yielded iso- 1992) and in the Ngorora Formation of the topic signature ranging from 5·2 to Tugen Hills (see Hill et al., 2002). Taxa 7·4‰ (Kingston et al., 1994) supporting shared with Maboko include Equatorius interpretations of a minor C4 grass compo- africanus. Nyanzapithecus, Victoriapithecus nent in the local ecosystem. These data macinnesi, Prodeinotherium cf. hobleyi, represent the oldest evidence for a C4 Protanancus macinnesi, Choerolophodon vegetation component in Africa, at about kisumuensis, possibly Paradiceros and 15·9 Ma. Kenyapotamus, Diamantohyus africanus, and Dorcatherium. And there are also a number Discussion of taxa at Kipsaramon not yet reported from Maboko. The fauna provides rather ambigu- Our dating of the Muruyur succession at ous clues regarding paleoenvironment. For Kipsaramon and at Muruyur itself provides example, it is not obvious how one recon- a tightly constrained chronostratigraphy for ciles the association of anomalurid flying the fossil-producing deposits, but also squirrels with peditid spring-hares, if the reveals problems with the placement of the habits of the Miocene species approximated Muruyur Beds in the regional stratigraphy. those of their extant counterparts. Accord- The Muruyur Beds originally were con- ing to Winkler (1992, 2002), anomalurids sidered to be a unit within the Kamuiton strongly indicate the presence of extensive Member of the Tiim Phonolites Formation, area of tropical forest. Peditids imply open but the set of dates from the lower part of habitats. Winkler (1990, 1992, 2002) also the succession presented here (Figure 3, comments in some detail on biogeographic Table 1) are outside the lower range of implications of the rodent sample. Stable previous Tiim dates and fit better with those carbon isotopic analyses of fossil enamel formerly obtained on the underlying Sidekh fragments of Rhinocerotidae and Pro- Phonolites Formation. However, dates on boscidea from BPRP#91 have produced phonolites overlying the unit at Muruyur are 13C values of 7·4 and 9·5‰ respect- consistent with other dates on upper Tiim ively (Kingston, 1992; Morgan et al., 1994). lavas. There is no inconsistency in the dating These data indicate that C4 grasses com- of the Muruyur Beds themselves, and prised a minor component of dietary intake although dates should not dictate stratigra- and were present in local ecosystems. C4 phy, it is clear that those reported here are plants (predominantly graminoids) are at odds with previous interpretations of the adapted to a warm-season growing period, regional stratigraphy. This problem has     35 broader importance as well, because these irregular topography without causing phonolite sequences are among the earliest additional erosion of the phonolite. volcanics in this part of the rift, and are Knobs or ridges on the surface of the relevant to the early history of the structure phonolite formed local embayments and its subsequent development (Chapman and topographic highs against which et al., 1978; King, 1970; King & Chapman, the sediments are banked. The paleo- 1972; Lippard, 1973). topography of the phonolite thus con- On a more local scale, the geological trolled local areas of emergence and history of Muruyur Beds at the Kipsaramon submergence on the margin of a lake site complex can be summarized as follows that opened out toward the east. (also see Figures 3, 5 and 6): ( 4) Deposition of silts, clays and lime- ( 1) Extrusion of a series of phonolitic lava stones indicating shifting subaerial and flows, with an edge of the final flow or subaquatic environment (subaerial red a fault scarp oriented approximately sediments with occasional paleosols, N–S in the Kipsaramon area and and subaqueous limestones and lami- downthrown toward the east. The nated green sediments; with all red Sidekh and Tiim phonolites represent deposits occurring below the level of an early expression of rift vulcanism in the BPRP#89 bone bed). Possible this part of East Africa, and previously interval of alkaline lacustrine con- documented thickening of the phono- ditions indicated by a zone of lites from the west of the Elgeyo Magadi-type chert nodules (Surdam Escarpment to the Tugen Hills in the & Eugster, 1976). The depositional east is interpreted as a consequence of environment throughout much of initial crustal sagging prior to rift fault- Member 1 appears to have been domi- ing (Chapman et al., 1978). Chapman nantly sub-aquatic, with influxes of (1971) also points out that faulting mudclast-rich coarse silts to sands and affected thicknesses of Muruyur reworked tuffs indicating proximity to deposits to the north, and it seems the lake margin. Periods of emergence likely that the lake basin(s) where are indicated by the oxidized pedo- these deposits accumulated was/were genic horizons, which are well devel- formed by an early phase of rift oped only in the north and northeast tectonics. part of the exposures. ( 2) Weathering of the flow surface and ( 5) Generally submergent and reducing emplacement of a phonolite agglomer- phase in southern part of area. Lake ate in the Gorge sectoin and also at the shallow near BPRP#89, poorly devel- section at Muruyur to the north. oped soils formed to north and ( 3) Soil formation on the upper phonolitic possibly to southwest. units, followed by erosion. The surface ( 6) Influxes of coarse, fluvially transported of the flow was stripped of most of its sediment (sand and mudclast con- weathered colluvium over much of the glomerates) including the bonebed Kipsaramon area prior to the begin- and discontinuous leaf mat of locality ning of the fluvio-lacustrine deposition BPRP#89. The fossiliferous unit was of Member 1. It appears that a chemi- distributed (not necessarily continu- cally and physically eroded volcanic ously) over a wide area (estimated at flow surface was transformed, rather between 0·09 and 0·28 km2). Based on abruptly, into a depositional site where the paleotopography of the phonolite water and sediment covered the and the lateral changes in Member 1 36 . .  ET AL.

lithofacies, the bonebed may have dates for the Muruyur sediments and for formed in a small embayment of a lake associated phonolites provide a high- basin that opened out toward the east, resolution chronostratigraphic framework with fluvial input from north to south for the unit and its contained fossil sites. suggested by the small paleochannel in The longest known succession is at Muruyur BPRP#89D as well as the presence of itself, with dates ranging from 15·98 a lake margin between BPRP#89G 0·07 Ma at the base to 13·370·06 Ma and F. near the top. In the region of Kipsaramon, ( 7) Lacustrine deposition over the whole on the crest of Tugen Hills, the top portion area; conditions generally not con- of Muruyur Beds is missing. The sediments ducive to benthic fauna and may have at Kipsaramon contain the most important been poorly oxygenated. fossil vertebrate sites. An extensive bonebed ( 8) Emplacement of Biryokwonin Tuff (BPRP#89) is one of the richest occurrences (ignimbrite), probably from the north. of in situ vertebrate remains in the East ( 9) Deposition of silts and reworked tuff African fossil record. Its age is constrained in a lacustrine environment with between 15·830·04 Ma and 15·59+ gastropods and bivalves preserved on 0·03 Ma. Another site (BPRP#122) has the very irregular surface of the produced a collection of the hominoid ignimbrite, followed by bedded tuffs, a Equatorius, dated between 15·58 Ma and layer of black chert, green clays, silty 15·36 Ma. Our preliminary interpretation of clays, and more finely bedded tuffs the bonebed accumulation is that it repre- (primary and reworked). sents a fluvial influx into a shallow lake. The (10) Swampy or shallow, partly oxygenated, fauna from the bonebed is taxonomically shallow lacustrine deposition, with clay comparable to others of a similar age in beds alternating with both reworked eastern Africa, such as that of Maboko. and primary ash fall tuffs (silt to sand- Anomalurid rodents in the bonebed fauna grade, with fresh, angular pumice frag- imply the presence of extensive tropical for- ments). This unit includes the bone est at the time, but peditids suggest some accumulation at BPRP#122. open areas. Isotopic analysis of sol carbon- The phonolites overlying the Muruyur ates indicate a minor component of C4 Beds are not present at the Kipsaramon site, grasses in the vegetation, the earliest such but this phase of vulcanism, which is well signal in the African record. documented farther north, presumably altered the paleogeography of the region and Acknowledgements may have ended deposition in the Muruyur lake basin. In any case, there is no higher This research forms part of the work of the sedimentary record available in the area of Baringo Paleontological Research Project the Kipsaramon site complex. (BPRP), based at Yale University, and car- ried out in collaboration with the National Museums of Kenya. We thank the Summary Government of the Republic of Kenya for The Muruyur Beds are an important sedi- permission to carry out research in Kenya mentary intercalation in the Middle (Permit OP/13/001/C 1391/ issued to AH), Miocene volcanic succession of the Rift and permission to excavate from the Valley. We have shown that their formal Minister for Home Affairs and National stratigraphic position in the regional succes- Heritage. BPRP has been supported by sion is uncertain. However, new radiometric grants to AH from NSF (most recently     37

SBR-9208903), the Louise H. and David S. Man, pp. 207–223. Edinburgh: Scottish Academic Ingalls Foundation, the Louise Brown Press. Chapman, G. R., Lippard, S. J. & Martyn, J. E. (1978). Foundation, Clayton Stephenson, and Yale The stratigraphy and structure of the Kamasia University. We thank Boniface Kimeu for Range, Kenya Rift Valley. J. Geol. Soc. Lond. 135, his considerable help with field and labora- 265–281. Deino, A. L. & Potts, R. (1990). Single-crystal 40Ar/ tory work, and the excavators of 1987 and 39Ar dating of the Olorgesailie Formation, Southern other field staff. The National Museums of Kenya Rift. J. geophys. Res. 95, 8453–8470. Kenya provided valuable logistic support, Deino, A. L., Tauxe, L., Monaghan, M. & Drake, R. (1990). Single-crystal 40Ar/39Ar ages and the litho- and we thank particularly Emma Mbua, and paleomagnetic stratigraphies of the Ngorora Collections Manager of Palaeoanthropol- Formation, Kenya. J. Geol. 98, 567–587. ogy, and Alfreda Ibui and Mary Mongu, Farquhar, G. D., O’Leary, M. H. & Berry, J. A. (1982). On the relationship between carbon isotope Collection Managers of Palaeontology. AH discrimination and the intercellular carbon dioxide thanks Rebecca Fisher for her help with concentration in leaves. Aust. J. Plant. Physiol. 9, faunal identifications that resulted in 121–137. Harrison, T. (1992). A reassessment of the taxonomic Table 2. We thank Peter Kat for the mollusc and phylogenetic affinities of the fossil catarrhines identifications presented in Table 3. AKB from Fort Ternan, Kenya. Primates 33, 501–522. thanks the NMNH Research Opportunity Hay, R. L. (1976). Geology of the Olduvai Gorge. Berkeley: University of California Press. Fund for support for field work in 1987. Jon Hill, A. (1989). Kipsaramon: a Miocene hominoid site Wingerath assisted with laboratory analyses in Kenya. Am. J. phys. Anthrop. 78, 241. of the sediments and also provided helpful Hill, A. (1994). Late Miocene and early Pliocene Hominoids from Africa. In (R. S. Corruccini & R. L. references. Ciochon, Eds) Integrative Paths to the Past: Paleo-  2002 US Government anthropological Advances in Honor of F. Clark Howell, pp. 123–145. Upper Saddle River, New Jersey: Prentice Hall. References Hill, A. (1995). Faunal and environmental change in the Neogene of east Africa: evidence from the Tugen Baker, B. H., Williams, L. A. J., Miller, J. A. & Fitch, Hills sequence, Baringo District, Kenya. In (E. S. F. J. (1971). Sequence and geochronology of the Vrba, G. H. Denton, T. C. Partridge & L. H. Kenya Rift volcanics. Tectonophysics 11, 191–215. Burckle, Eds) Paleoclimate and Evolution, with Baker, B. H., Mohr, P. A. & Williams, L. A. J. (1972). Emphasis on Human Origins, pp. 178–193. New Geology of the Eastern Rift System of Africa. Geol. Haven: Yale University Press. Soc. Am. Sp. Paper 136, 1–67. Hill, A. (1999). The Baringo Basin, Kenya: from Bill Best, M. G., Christiansen, E. H., Deino, A. L., Bishop to BPRP. In (P. Andrews & P. Banham, Eds) Gromme´, C. S. & Tingey, D. G. (1995). Emplace- Late Cenozoic Environments and Hominid Evolution: a ment of the large volume, rhyolitic Pahranagat Tribute to Bill Bishop, pp. 85–97. London: Geological ash-flow tuff, Nevada: 40Ar/39Ar chronology, paleo- Society of London. magnetism, and petrology. J. geophys. Res. 100, Hill, A. & Ward, S. (1988). Origin of the Hominidae: 24593–24609. the record of African large hominoid evolution Bishop, W. W., Miller, J. A. & Fitch, F. J. (1969). New between 14 My and 4 My. Yearb. phys. Anthrop. 31, potassium-argon age determinations relevant to the 49–83. Miocene fossil mammal sequence in east Africa. Am. Hill, A., Curtis, G. & Drake, R. (1986). Sedimentary J. Sci. 267, 669–699. stratigraphy of the Tugen Hills, Baringo, Kenya. Bishop, W. W., Chapman, G. R., Hill, A. & Miller, In (L. E. Frostick, R. W. Renaut, I. Reid & J.-J. J. A. (1971). Succession of Cainozoic vertebrate Tiercelin, Eds) Sedimentation in the African Rifts, assemblages from the northern Kenya Rift Valley. pp. 285–295. Oxford: Geological Society of London Nature 233, 389–394. Special Publication No. 25. Brown, B., Hill, A. & Ward, S. (1991). New Miocene Hill, A., Behrensmeyer, A. K., Brown, B., Deino, A., large hominoids from the Tugen Hills, Baringo Rose, M., Saunders, J., Ward, S. & Winkler, A. District, Kenya. Am. J. phys. Anthrop. 12, 55. (1991). Kipsaramon: a lower Miocene hominoid site Chapman, G. R. (1971). The geological evolution of in the Tugen Hills, Baringo District, Kenya. J. hum. the northern Kamasia Hills, Baringo District, Kenya. Evol. 20, 67–75. Ph.D. Dissertation, University of London. Hill, A., Leakey, M., Kingston, J. & Ward, S. (2002). Chapman, G. R. & Brook, M. (1978). Chronostra- New cercopithecoids and a hominoid from 12·5 Ma tigraphy of the Baringo Basin, Kenya Rift Valley. In in the Tugen Hills succession, Kenya. J. hum. Evol. (W. W. Bishop, Ed.) Geological Background to Fossil 1/2, 75–93. 38 . .  ET AL.

Kelley, J., Ward, S., Brown, B., Hill, A. & Downs, W. Renne, P. R., Swisheer, C. C., Deino, A. L., Karner, (2000). Middle Miocene hominoid origins. Science D. B., Owens, T. L. & dePaolo, D. J. (1998). 287, 2375. http://www.sciencemag.org/cgi/content/ Intercalibration of standards, absolute ages and full/287-5462/2375a. uncertainties in 40Ar/39Ar dating. Chem. Geol. 145, Kelley, J., Ward, S., Brown, B., Hill, A. & Duren, D. 117–152. (2002). Dental remains of Equatorius africanus from Retallack, G. J. (1991). Miocene Paleosols and Ape Kipsaramon, Tugen Hills, Baringo District, Kenya. Habitats of Pakistan and Kenya. New York: Oxford J. hum. Evol. 1/2, 39–62. University Press. King, B. C. (1978). Structural and volcanic evolution Sherwood, R. J., Ward, S. C., Hill, A., Duren, D., of the Gregory Rift Valley. In (W. W. Bishop, Ed.) Brown, B. & Downs, W. (2002). Preliminary Geological Background to Fossil Man, pp. 29–54. description of the Equatorius africanus partial skeleton Edinburgh: Scottish Academic Press. (KNM-TH 28860) from Kipsaramon, Tugen King, B. C. & Chapman, G. C. (1972). Volcanism of Hills, Baringo District, Kenya. J. hum. Evol. 1/2, the Kenya rift valley. Phil. Trans. R. Soc., Lond. 271, 63–73. 185–208. Singer, A. & Stoffers, P. (1980). Clay mineral diagen- Kingston, J. D. (1992). Stable isotopic evidence for esis in two East African lake sediments. Clay Minerals hominid paleoenvironments in East Africa. Ph.D. 15, 291–307. dissertation, Harvard University. Surdam, R. C. & Eugster, H. P. (1976). Mineral Kingston, J. D., Marino, B. D. & Hill, A. (1994). reactions in the sedimentary deposits of the Lake Isotopic evidence for Neogene hominid paleo- Magadi region. Bull. Geol. Soc. Am. 87, 1739–1752. environments in the Kenya Rift Valley. Science 264, Ward, S., Brown, B. & Hill, A. (1996). Forelimb 955–959. of Kenyapithecus africanus from the Tugen Hills, Lippard, S. J. (1973). Plateau phonolite lava flows, Baringo District, Kenya. Am. J. phys. Anthrop. Kenya. Geol. Mag. 110, 543–549. 22(Suppl.), 240. Martyn, J. (1969). The geologic/history of the country between Lake Baringo and the Kerio River, Baringo Ward, S., Brown, B., Hill, A., Kelley, J. & Downs, W. District, Kenya. Ph.D. dissertation, University of (1999). Equatorius: a new hominoid genus from London. the Middle Miocene of Kenya. Science 285, 1382– Morgan, M. E., Kingston, J. D. & Marino, B. D. 1386. (1994). Carbon isotopic evidence for the emergence Winkler, A. J. (1990). Systematics and biostratigraphy of C4 plants in the Neogene from Pakistan and of Neogene rodents from the Baringo District, Kenya. Nature 367, 162–165. Kenya. Ph.D. dissertation, Southern Methodist Ness, G., Levi, S. & Crouch, R. (1980). Marine University. magnetic anomaly time scales for the Cenozoic and Winkler, A. J. (1992). Systematics and biogeography late : a pre´cis, critique, and synthesis. Rev. of middle Miocene rodents from the Muruyur Geophys. Space Phys. 18, 753–770. Beds, Baringo District, Kenya. J. Vert. Paleontol. 12, O’Leary, M. O. (1988). Carbon isotopes in photo- 236–249. synthesis. BioScience 38(5), 328–336. Winkler, A. J. (1994). The Middle/Upper Miocene Pickford, M. (1975). Stratigraphy and palaeoecology of dispersal of major rodent groups between southern five Cainozoic Formations in the northern Kenya Rift Asia and Africa. In (Y. Tomida, C. K. Li & T. Valley. Ph.D. dissertation, University of London. Setoguchi, Eds) Rodent and Lagomorph Families of Pickford, M. (1988). Geology and fauna of the middle Asian Origins and Diversification, pp. 173–184. Miocene hominoid site at Muruyur, Baringo district, Tokyo: National Science Museum Monographs, Kenya. Hum. Evol. 3, 381–390. Vol. 8. Plummer, T. W., Kinyua, A. M. & Potts, R. (1994). Winkler, A. J. (2002). Neogene paleobiogeography and Provenancing of hominid and mammalian fossils East African paleoenvironments: contributions from from Kanjera, using EDXRF. J. Archaeol. Sci. 21, the Tugen Hills rodents and lagomorphs. J. hum. 553–453. Evol. 1/2, 237–256.