JOURNAL OF THE ROYAL SOCIETY OF WESTERN AUSTRALIA VOLUME 61 PART 3 JANUARY, 1979. PRICE TWO DOLLARS REGISTERED FOR POSTING AS A PERIODICAL-CATEGORY 8 THE ROYAL SOCIETY OF WESTERN AUSTRALIA PATRON Her Majesty the Queen VICE-PATRON His Excellency Air Chief Marshal Sir Wallace Kyle, G.C.B., K.C.V.O., C.B.E., D.S.O., D.F.C., K.St.J., Governor of Western Australia President Vice-Presidents .. Past President Joint Hon. Secretaries Hon. Treasurer Hon. Librarian Hon. Editor COUNCIL 1978-1979 . C. F. H. Jenkins, M.B.E., M.A., M.A.I.A.S. . M. J. Mulcahy, B.Sc. (For.), Ph.D. J. R. de Laeter, B.Ed.(Hons.), B.Sc. (Hons.), Ph.D., F.Inst.P., F.A.I.P. . A. J. McComb, M.Sc., Ph.D. G. Perry, B.Sc. (Hons.) M. W. Perry, B.Sc. (Agric.) (Hons.) S. J. Curry, M.A. A. Neumann, B.A. A. E. Cockbain, B.Sc., Ph.D. B. J. Beggs, B.Sc. (For.), Dip. For. B. K. Bowen, B.Sc. W. C. Ferguson, B.A. (Hons.) J. K. Marshall, B.Sc. (Hons.), Ph.D. L. J. Peet, B.Sc., Dip.Val., Dip.R.E.M., F.G.S., A.R.E.I. P. E. Playford, B.Sc., Ph.D. J. C. Taylor, B.Sc., Ph.D., A.R.C.S. P. R. Wycherley, O.B.E., B.Sc., Ph.D., F.L.S. Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979, p. 67-72. The Whim Creek Group, a discussion by R. C. Horwitz CSIRO, Division of Mineralogy, Floreat Park, W.A. 6014 Manuscript received 21 February 1978; accepted 16 May 1978 Abstract The Mallina Formation and the Constantine Sandstone should be placed in the Whim Creek Group, following the proposal of Fitton et al. (1975), and not in the underlying Gorge Creek Group as advocated by Hickman and Lipple (1975) and Hickman G977). BIF in the Gorge Creek Group occurs above, but separated from, fuchsite-bearing metasediments; both lithologies are useful marker bands. Their distribution as fragments in the Constantine Sandstone gives additional data regarding the level of erosion that preceded the deposition of the Whim Creek Group. The Whim Creek Group consists of a predominantly clastic province and a predominantly volcanic province, separated, during sedimentation, by a hinge zone, which was the site of later emplacement of quench-textured rocks. Introduction The Whim Creek Group has been described and defined by Fitton et al. (1975). The group occurs in the West Pilbara region of Western Australia, and rests unconformably on the Gorge Creek Group and the Teichmans Group, both of older Archaean age. These two underlying units are correlated respectively with the Soansville Sub-group (essentially) and the Warrawoona Group as defined in the eastern part of the Pil- bara Block by Hickman and Lipple (1975). An account of relevant studies and of correlations with units of the eastern half of the Pilbara Block is given by Fitton et al. (1975). The Whim Creek Group consists mainly of volcanic rocks in the Mons Cupri area and meta- sediments to the southeast, referred to respec- tively as a volcanic province and a clastic pro- vince by Horwitz and Smith (1978). The tectonic setting and relationship between these two provinces is illustrated in Fitton et al. (1975, Figs. 1, 2. I record here an error in Fig. 1; the Gorge Creek Group does not extend west of Loudens Fault, near Peawah Hill). Volcanics and sediments are considered to intertongue and to equate in time for all but the highest units in the clastic sequence. The type area for several of the volcanic members is defined at Mons Cupri, and the formations of the clastic province are equated to particular units of this type section (Fitton et al. 1975, p. 17). The Whim Creek Group comprises the Warambie Basalt, Mons Cupri Volcanics, Con- stantine Sandstone and Mallina Formation; the formations are not well-developed everywhere, and in some places are absent. The Negri Vol- canics have been excluded as they overlie the sequence, unconformably in places. 77266 — ( 2 ) Hickman (1977) presents a map of part of the volcanic province which he names the Whim Creek Belt. Based on a relationship between porphyritic rocks and metasediments, 15 km southeast of Sherlock, as well as on an interpre- tation of structural data, he concludes (p. 56) : “The mid-Archaean regional unconformity recognised by Fitton and others (1975) has not been substantiated by regional mapping . . . The Mallina Formation and Constantine Sand- stone, placed by Fitton and others (1975) in the Whim Creek Group, belong to the Gorge Creek Group”. This paper presents new syntheses based on published data in support of Fitton et aVs. thesis. For references to localities and to the distribu- tion of rock units, the reader is referred to Fitton et al. (1975, Fig. 1), Hickman (1977, Fig. 28), and the 1:250 000 geological maps (Roe- bourne and Pyramid sheets) published by the Geological Survey of Western Australia. The Mons Cupri Volcanic Province The Mons Cupri Volcanic Province is used in preference to the term “Whim Creek Volcanic Belt” so that one can include, within this pro- vince of volcanic rocks, those acid volcanics, tuffs and sediments which are preserved in the syncline between the Caines Well and the Balia Balia Granites (which are cut by Salty Creek and Balia Balia River), as well as the remnants, large rafts and roof pendants, in the gabbros and granophyres of the Millindinna Complex and in the possibly related granites, which occur between the George and Little Sherlock Rivers. The basal unit of the Whim Creek Group in the Mons Cupri Volcanic Province is the Waram- bie Basalt which occurs wherever the basal con- tact is not intruded by units of the Millindinna Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. Complex. The basalt is more widespread than shown by Hickman (1977) between the Sherlock River and the road from Sherlock to Croydon. The Warambie Basalt is characteristically pitted by numerous small vesicles, recognisable even where the rock is sheared or altered. This, as well as their stratigraphic position, justifies including in the Warambie Basalt (as in Fitton et al., Fig. 1) the basic rocks which occur between 2 and 5 km northeast of Mons Cupri. Pebbles of Warambie Basalt occur in elastics and in vol- canic fragmentals in many exposures in the volcanic province of the Whim Creek Group. The basal unconformity (Fitton et al. 1975, Hickman 1977) outcrops along the old road to Pyramid between Black Hill and Red Hill (be- tween the George and Little Sherlock Rivers). There, the Warambie Basalt is nearly horizontal and rests on basic rocks in which regional folds plunge to the northwest. These are ascribed to the Teichmans Group. The unconformity is emphasised by the contrasting grades of meta- morphism, estimated to be low in the overlying rocks which appear in hand specimen to be scarcely altered, whereas the underlying basic rocks appear to be amphibolite-grade rocks; the contained amphiboles are dark in colour and the rocks are strongly lineated. To the south, the Warambie Basalt is overlain by coarse clastic rocks and volcanic fragmentals (dolerite or gabbro intrudes the contact), and the dips steepen southwards. The Warambie Basalt here is therefore on the crest of the south-facing limb of an anticline. Figure 1 is a geological cross-section at an azimuth of 210°, passing through the mine at Mons Cupri, approximately through Whim Creek Mine, and extending southwestward to 10 km from Mons Cupri. The deepest unit is the Mount Brown “Rhyolite Member”. As suggested by Fitton et al. (1975, p. 16) and confirmed by later work (G. Sylvester, pers. comm. 1976), the unit is believed to be a chilled intrusive rock. Indeed, it contains many small xenoliths of country rock, particularly abundant close to the contacts. This unit forms a broad dome, intrud- ing phyllites to the northeast and volcanic frag- mentals to the south. Similar rocks occur else- where in the province. The volcanic fragmentals (“Mons Cupri rhyolite fragmental”) have been described by Miller & Gair (1975) and are now considered to be the oldest unit exposed in the section (Fig. 1). At the Mons Cupri mine the frag- mentals are chloritized in places. This change masks the fragmental appearance but it becomes conspicuous again where the rocks are weathered. Mapping of the cleavage and the predominant orientation of the long axes of the fragments established that the fragmental unit is wedge- shaped, thinning rapidly towards the dome of the Mount Brown Member, but thinning more gradually southwards. Thus the volcanic fissure, or vent, responsible for the accumulation of the fragmental unit, was probably nearby to the north, and may have controlled the later em- placement of the intrusion of the Mount Brown Member. This general convergence of primary trends towards the north is reflected by the shape of the net-veined ore-body. Sandstones, grits and conglomerates (Cistern Formation, Miller & Gair 1975) overlie the volcanic fragmentals. This unit correlated with the Constantine Sandstone by Fitton et al. (1975, p. 17) has caused much geological debate; it does intrude, or mix with, the underlying fragmentals, but on petrological examination it is essentially a tuffaceous grit. The bulk of the rock is massive, although graded bedding has been observed in drill core (K. O. Linn, pers. comm. 1972). At the top, it contains well defined layers, lenses and pockets of fine-grained material or boulder beds. Volcanic pebbles and fragments of fuchsite schist occur in places. Phyllites overlie this unit and are equated (Fig. 1) with the phyllitic slate at Whim Creek and with the metasediments below the Negri Volcanics to the southwest. Near Mons Cupri, a band of basic rock occurs close to the base of the Mallina Formation. It is too thin to depict in Figure 1. The rock is strongly altered and structureless, apart from some pronounced near- vertical jointing; its con- tacts with the phyllites are covered by rubble. The unit was named the Comstock Andesite by Miller & Gair (1975) but lacks volcanic criteria and is probably a fine-grained basic intrusive. Similar fine-grained basic rocks occur elsewhere NEGRI VOLCANICS MOUNT BROWN MEMBER (intrusive) Bedded phyllites with gritty and tuffaceous (fragmental) bands. MALLINA FORMATION. Sandstones (tuffaceous), grits and conglomerates. Volcanic fragmental, with thin, rare, sedimentary tuffaceous intercalations. Fault SCALE Ver tical _ . Horizontal topography is exaggerated 0 2 km Figure 1.— Geological cross-section, striking 210°, through Mons Cupri Mine. 68 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. in the metasediments throughout the volcanic province. This does not, however, exclude the possibility of basic flows occurring with the sediments; indeed M. J. Fitton (pers. comm. 1977) has mapped possible lavas (a fine-grained basic rock) in the Mallina Formation near the Peawah River at Egina, in the clastic province. Two areas of metasediments shown on Figure 1, flanking the Mount Brown Dome, are mapped by Hickman (1977, Fig. 28) as “slate (e.g. at Whim Creek)’’ and excluded from the Mallina Formation, whilst a third area, on the south limb of the syncline with Negri Volcanics, is included in the Mallina Formation. There is no evidence to indicate a discontinuity beneath this syncline. Hickman agrees (p. 55) that “the slate at Whim Creek lithologically resembles more pelitic parts of the Mallina Formation’’; how- ever, he has found that the sediments, beyond the southwest extension of Figure 1, dip under a porphyritic unit which he has included in the Mons Cupri Volcanics. These outcrops were accidentally omitted from the map in Fitton et al. (1975, Fig. 1). However, they do not con- tradict the conclusions of Fitton et al. regarding the extent, or unconformable nature of the Whim Creek Group. Such porphyries, some of which are flow-banded in places, are widespread throughout the volcanic province and were in- cluded by Fitton et al. as an integral part of the Mons Cupri Volcanics. Similar rocks have since been recorded in the Mallina Formation near its base, between Mt. Satirist and Millindinna by M. J. Fitton (pers. comm. 1977) and also about 10 km north of Egina by G. Doust (pers. comm. 1977), thus extending their distribution to the clastic province. Palaeogeographic evolution of the Archaean in the West Pilbara Early mapping in the West Pilbara by Kriewaldt et al. (in Kriewaldt 1964) established a persistent sequence, applicable to the Archaean of the Pilbara Block west of Roebourne. The sequence, briefly described by Ryan and Kriewaldt (1963), was later abandoned (Ryan and Kriewaldt 1964, Ryan 1965) for the region between Mons Cupri and Mt. Satirist, following a misinterpretation of the relative ages of the later named Croydon Sandstone and the Gorge Creek Group (see Fitton et al. 1975, p. 5-6). The seauence from top to bottom (numbered as in Fig. 2) is, (4) Gorge Creek Group, BIF and sediments; (5) basic to intermediate volcanics, frequently pillowed; (6) a composite assemblage of mafic to ultra-mafic rocks with acid volcanics and sediments (Nickol River Formation of Williams 1968). Pillowed basalt underlies unit (6), where preserved by granitic intrusion. The granitoids intrude all units below the Gorge Creek Group. Unit (6) is characterised by frequent occur- rences of green fuchsite (a chromian muscovite) in the sedimentary bands. It is a complex stratigraphic unit (see Horwitz 1963). Correla- tions with the type section in the Teichmans region, where a similar sequence occurs, are shown in Figure 2; they were established with the help of M. J. Fitton. These chrome-bearing sediments, whose geochemistry still remains to be studied, could be genetically related to ultra- mafic volcanicity which, where developed, occurs in the same general part of the sequence. Sherlock ? ? Toweranna -fTTT+ + + + + 4- + + + + + + + + + + + + +++++++++ 4- + + + + + + + + + + + + + + 4- 4- 4- + 4- 4- ‘ -T yT 4- 4- H -*--+--+--*- 4 --*- + + + + + + + + + + + + + + + 4; + + + + + + + + + + + + + + + + + + + + + + ++ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + ' , t' + 4 B iii m i n <• 4- + + + + + + + + + + + + + + + + + + + + + + + + ++ + + + + ++++++++++++++++++++++++++++++++++++++++++++++++++++ 4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4- + 4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4- Sea level +++++++++++++ +++++++++++++++++++++++++ +++++++++++++++++++++++++++++++ +++++++++++++++++++++++++++++++ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +~ : F^ r + + + + + + + + + + + + + + + + + + ++ + + + + + + + + + + + + + + + + + + + + + + + + + + *-+ + + + + + + + + + + + + + + + + + + + + + + + + + + + 4 +++++++++4 f +++++++++4 WHIM CREEK GROUP TEICHMANS GROUP Granitic rocks WYMAN FORMATION EMPRESS FORMATION FRIENDLY CREEK FORMATION 1 MALLINA FORMATION 2 CONSTANTINE SANDSTONE 3 MONS CUPRI VOLCANICS 4 GORGE CREEK GROUP 5 Volcanic rocks 6 Contains cherts and fine grained sediments, in part fuchsitic SCALE Vertical _ ^ Horizontal 0 10 20 Hu, Figure 2. — Three diagrammatic palaeogeographic profiles approximately through Roebourne and Wodgina. A —Fol- lowing deposition of the Gorge Creek Group and intrusion by granitic rocks. B.— Prior to deposition of the Whim Creek Group. C. — After deposition of the Whim Creek Group. 69 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. The Gorge Creek Group BIFs, and the cherts and other sediments with the green fuchsite staining, thus provide two broad marker units which have been used to determine the level of erosion prior to deposition of the Whim Creek Group. These data have been used in associa- tion with the level of intrusion of the granites and the thicknesses and facies of the late Archaean rocks of the region in the compilation of the palaeogeographic profiles in Figure 2 which is based on the profiles in Fitton et al. (1975, Fig. 2). Rare granitoid pebbles, as well as fuchsitic fragments, occur in the volcanic fragmentals at Mons Cupri, and fuchsitic fragments occur in the clastic metasediments correlated with the Constantine Sandstone at Mons Cupri. They are also present in the Constantine Sandstone of the Croydon Anticline, and the Mallina Anti- cline south of Loudens Fault. M. J. Fitton has recorded some, associated with fragments derived from the Gorge Creek Group, in the Constantine Sandstone north of Teichmans Goldmine. One can recognise a probable source for all of these clasts, including the fuchsitic fragments, which almost certainly derive from unit 6. In most other places, throughout the Teichmans and Mt. Satirist general area, the Constantine Sand- stone contains chert and BIF pebbles derived from the Gorge Creek Group (Fig. 3). Such pebbles are usually well rounded. Very large slabs of chert and BIF occur at the base of the Constantine Sandstone north of Teichmans Goldmine; and in other areas the basal forma- tion is a ferruginous breccia (M. J. Fitton, pers. comm. 1975). All this, when allied to an irregu- larity in detail of the surface of unconformity, indicates that the Gorge Creek Group was well lithified and eroded prior to the deposition of the Whim Creek Group. Where the Constantine Sandstone is very thin and reduced to a few metres of elastics, it has been omitted from Figure 2C. The profiles in Figure 2 depict three stages in the evolution of the West Pilbara. Figure 2 A is a profile after deposition of the Gorge Creek Group and following intrusion by granitic rocks. Nowhere in the West Pilbara are the fuchsitic sediments (6) in normal contact with the BIF (4) ; they are always separated by volcanic rocks. Figure 2B indicates a broad arching of the sequence in the general area where granitic rocks have reached overall higher levels of intrusion. This might support the suggestion of M. J. Fitton (in Fitton et al. 1975) that the diapiric rise of granitoids is responsible for the structural deformation in the region. It is possible that this tectonism started before com- plete lithification of the Gorge Creek Group BIF as the latter show remarkable plastic deforma- tion at both small and large scales. However, Figure 3. — Conglomerate from the Constantine Sandstone at Nunyerry Gap. The pebbles are largely of Gorge Creek Group BIF. Sample provided by M. J. Fitton. 70 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. diapirism, if it occurred, is considered to have been accompanied by clear intrusion with stop- ing and the injection of large granite sills and apophyses, as described by Horwitz (in Fitton et al. 1975, p. 19) for the Caines Well Granite. Indeed, the zone of mafic-ultramafic remnants, which was depicted in the granite from aero- magnetic maps, does outcrop in places as small and large remnants. The zone can be traced from the outcrops at Sherlock, southwestwards to an outcrop, adjoining and north of the new highway on the left bank of the Little Sherlock River (recorded in Fitton et al. 1975). Further westwards, the zone is displaced by the Copper Mine Fault, as the next outcrop is a small mound of granophyre of the Millindinna Com- plex (brecciated according to G. H. Riley, pers. comm.). The trace of the Copper Mine Fault, as depicted in Fitton et al. (1975) is confirmed by aeromagnetic data. These remnants in the granites, as well as the country rocks in contact with the intrusive granites, are invariably metamorphosed to a high grade. It is to this contact effect that we attribute the metamorphic grade of the Teich- mans Group amphibolites, below the Warambie Basalt. The contact effect also explains the previously unrecorded metamorphosed mafic- ultramafic rocks, preserved and sandwiched be- tween the Millindinna Complex and the Croydon Sandstone at the Evelyn Copper Mine near Croydon (see also, Williams 1968, p. 7). Where the dips are shallow the Mallina For- mation phyllites have a pronounced cleavage resulting from flexure slip folding and sub- parallel to, but not to be confused with, the bedding. This occurs, for example, at Whim Creek, on both limbs of the Croydon Anticline a few kilometres away from the fold axis, near Egina, and northwest of Teichmans Goldmine. However, the steeper fracturing, or axial-plane cleavage is developed wherever the bedding is steeper, such as in the Mallina Formation near Mt. Negri, below the unconformity of the Negri Volcanics. Thus the steepness of the beds is not a criterion of age, as implied by Hickman (1977, p. 55). Hickman (1977) has separated from the Negri Volcanics, his unit “Abu”, the quench-textured basic rocks. West of the area covered by his map, close to and west of Warambie Homestead, the unit appears to intrude between the Waram- bie Basalt and other units of the Whim Creek Group. Its geographic distribution, largely flanking the volcanic province, (compare with Figure 20 suggests that these mafic rocks are developed mainly at the hinge zone of thickness variations, which separates the volcanic pro- vince from the clastic province in the Whim Creek Group. As noted by Hickman (1977), this zone is marked by faults, the major one being Loudens Fault. These features could sug- gest that some deeper crustal control was possibly responsible for many aspects of the palaeogeography and mineralisation in the region. Hickman (1977, Fig. 28), however, erred in separating these quench-textured rocks from others of the Negri Volcanics, such as those that overlie the Whim Creek Group in the syncline about 10 km south of Sherlock Homestead. In- deed, Hallberg (1973, p. 6) has noted similarities in textures and chemical affinities between his samples “Sherlock” and his samples “Mt. Negri Volcanics”. The latter are from Hickman’s unit “Abu” and the former from his “silicified and epidotised basalt”. Accepting that these two units are part of the same uninterrupted sequence, as originally mapped by Fitton et al. (1975), then the quench- textured basic rocks are unquestionably younger than the Whim Creek Group and cannot corre- late with others of the Teichmans Group as suggested by Hickman (1977, p. 56). Indeed both Hallberg (1973) and Sun & Nesbitt (1978, p. 316) give chemical evidence showing that these quench-textured rocks of the Negri Vol- canics are more differentiated than those com- mon to the Archaean (for instance, Ruth Well, Hallberg 1973, Table 3). Conclusions The Constantine Sandstone and the Mallina Formation are part of a sequence which is con- sidered to be unconformable on, and thus excluded from, the Gorge Creek Group of BIF. The deposition of the Constantine Sandstone followed a period of erosion and folding; dissec- tion must have proceeded at least as deep as unit (6). Thus the Whim Creek Group is a valid unit, which includes both volcanic and sedi- mentary rocks, and is distributed quite widely throughout the Archaean of the Pilbara Block. Its extent is indicated and discussed in Fitton et al. (1975, Fig. 3 and p. 23). The relative ages of the Millindinna Complex, of some granitic rocks and of various rocks grouped with the Negri Volcanics, deserves more research. The age of some of these units might be a matter of semantics, to be solved by geo- chronological dating and an agreed definition of the boundary between the Archaean and the Proterozoic. Horwitz and Smith (1976, 1978) have shown that volcanicity was nearly con- tinuous from late Archaean to early Proterozoic in parts of an early Proterozoic trough, super- imposed on the area discussed in this paper. Indeed, there is still ample scope for further research in the region. a cKrujivLeagemenzs . — me concepts outlined here are essentially those developed with M. J. Fitton and G Sylvester, during research carried out prior to our •jomt paper in 1975. My thanks to all those acknow- ledged in Fitton et al. are here renewed and the opportunity is taken to repair the following omission- the comments by Anhaeusser (1971, d. 117), although pertinent to the unconformity of the Whim Creek Group, were not included in the historical section (Fitton et al. 1975, p. 2-9), because they were noted too late for publication. Since publication in 1975 my thoughts on the region have been clarified bv discussion with and information supplied by M. j Fitton and G. Sylvester and by discussions at the Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. outcrops with R. Carey, G. Doust, A. Y. Glikson, R. C. Morris, A. S. Novikova, D. Philp, G. H. Riley and Sv. A. Sidorenko. The help, in recent discussions of J. A. Hallberg and E. S. T. O’Driscoll is also recorded. I thank both Mr. W. E. Ewers and Dr. E. H. Nickel for critically reading the manuscript and Mr. C. R. Steel for drafting the figures. References Anhaeusser, C. R. (1971).— The Barberton Mountain Land. South Africa — A Guide to the Under- standing of the Archaean Geology of West- ern Australia. In Glover, J. E. (edit.). Symposium on Archaean rocks held at Perth. 23-26 May 1970. Geol. Soc. Australia Special Publication 3: 103-119. Fitton, M. J., Horwitz, R. C. and Sylvester, G. ( 1975) . — Stratigraphy of the Early Precambrian in the West Pilbara, Western Australia. Aus- tralia Commonwealth Sci. Industrial Re- search Organization, Minerals Res. Labs. Rept. FP. 11, 31 p. Hallberg, J. A. (1973). — Whole rock geochemical orien- tation trip to the Pilbara. Australia Commonwealth Sci. Industrial Research Organization, Minerals Res. Labs. Rept. FP. 3, 12 p. Hickman, A. H. (1977). — Stratigraphic relations of rocks within the Whim Creek Belt. Geol. Surv. West. Australia , Ann. Rept. 1976: 53-56. Hickman, A. H. and Lipple, S. L. (1975). — Sheet Marble Bar. Geol Surv. West. Australia. Rec. 1974/20. Horwitz, R. C. (1963). — Facies changes in the Archaean of the Roebourne area, West Pilbara Gold- field. Geol. Surv. West. Australia, Ann. Rept. 1962: 37. Horwitz, R. C. and Smith, R. E. (1976). — Bridging the Yilgarn and Pilbara Blocks, Western Aus- tralian Shield. Section 1A, No. 9, 25th International Geol. Congress, Sydney, August 1976. (Unpublished address). Horwitz, R. C. and Smith, R. E. ( 1978) .—Bridg- ing the Yilgarn and Pilbara Blocks, Western Australia. Precambrian Res. 6: 293-322. Kriewaldt, M. (1964). — Dampier and Barrow Island, Western Australia. 1:250 000 Geol. Series Explanatory Notes. Geol. Surv. West. Aus- tralia, 13 p. Miller, L. J. and Gair, H. S. (1975).— Mons Cupri copper-zinc-lead deposit. In Knight, C. C. (edit.). Economic Geology of Australia and Papua New Guinea — Metals. Austra- lasian Inst. Mining Metall. Mon. 5: 195-202. Ryan, G. R. (1965). — The geology of the Pilbara Block. Australasian Inst. Mining Metall. Proceed. 214: 61-94. Ryan, G. R. and Kriewaldt, M. (1963). — Archaean stratigraphy in the Roebourne area, West Pilbara Goldfield. Geol. Surv. West. Austra- lia, Ann. Rept. 1962: 38. Ryan, G. R. and Kriewaldt, M. (1964). — Facies changes in the Archaean of the West Pilbara Gold- field. Geol. Surv. West. Australia, Ann. Rept. 1963: 28-30. Sun, Shen-Su and Nesbitt, R. W. (1978). — Petrogenesis of Archaean ultrabasic and basic volcanics: Evidence from rare earth elements. Contrib. Mineral. Petrol. 65: 301-325. Williams, I. R. (1968). — Yarraloola, Western Australia. 1:250 000 Geol. Series Explanatory notes. Australian Bur. Min. Res. 30 p. 72 Journal of the Royal Society of Western Austrlia, Vol. 61, Part 3, 1979, p. 73-96. Prehistoric rock wallabies (Marsupialia, Macropodidae, Petrogale) in the far south-west of Western Australia by D. Merrilees Western Australian Museum, Francis St., Perth, W.A. 6000 Manuscript received 21 March 1978; accepted 18 April 1978 Abstract Rock wallaby remains are described from several caves in the Cape Leeuwin- Cape Naturaliste region. Measurements of cheek teeth reveal variations in size from place to place and to some extent from time to time, but there do not seem to be any accompanying differences in form. Hence the samples are regarded as conspecific, probably representing Petrogale penicillata. It is suggested that rock wallabies arrived in the region prior to 30 000 yr B.P. and disappeared well before historic time, the disappearance perhaps being due to replacement of a more open vegetation by dense forest in early Holocene time. In an appendix, there is a record of the arrival of the dog prior to the local extinction of the thylacine, and in another appendix, the presence of Lagorchestes in the region is confirmed. Introduction Rock wallaby specimens had been found in various parts of south-western Australia, includ- ing the Cape Leeuwin-Cape Naturaliste region covered in this paper, in the early years of the present century, but had not been recognized as such, and had been stored in the Western Aus- tralian Museum collection with other macropods of similar size, notably Macropus irma and Setonix brachyurus. The earliest published reference to them in the Cape Leeuwin-Cape Naturaliste region appears to be that of Lundelius (I960) reporting a discovery made in 1955 in what is now called Devil’s Lair. Even after this discovery, Petrogale was still confused with other taxa in the Museum collection. Examples are given below. My attention was first drawn to this confusion in 1965, in connec- tion with a large collection of very juvenile Petrogale specimens made by D. L. Cook in 1958 in Deepdene Cave. The present paper attempts to resolve this confusion and to review the occurrence of rock wallaby remains in a region from which they were conspicuously absent in historic time (Calaby 1971), but a notable part of the macro- pod fauna in prehistoric time (Baynes et al. 1976). I have also taken the opportunity to review some intrinsically interesting deposits which happen to include Petrogale, and some samples of Petrogale which, though restricted in character, are more extensive than those likely to be available to neontologists, and which repre- sent a range in time as well as in space. The report covers all Petrogale specimens from the Cape Leeuwin-Cape Naturaliste region in the Western Australian Museum collection at the time of writing (January 1978). It is not practicable to list their catalogue numbers, but a copy of raw data on specimens measured, including their catalogue numbers, has been lodged in the Museum library, and particular specimens are cited by number in context. The localities from which Petrogale specimens have been derived are described first, beginning with the most southerly (The Labyrinth) and proceeding to the most northerly (Yallingup Cave), about 70km away. Cave numbers are those listed by Bridge (1972, 1973) and Bridge and Shoosmith (1975). Many of the sites are mapped by Lowry (1967) in describing the general geology of the region. The locality des- criptions are followed by descriptions of the samples, and these in turn by some discussion of their significance. The present day vegetation of the region is described by Smith (1973), but there may be considerable differences between the vegetation in historic time, and that of the prehistoric period during which Petrogale flourished (Balme et. al. 1978, Churchill 1968). Localities The Labyrinth (AU16). — The general form of this aptly named cave, about equidistant from the townships of Augusta and Karridale, near Cape Hamelin, is described by Lowry and Bain (1964). 73 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. Remains of extinct marsupials were found by D. C. and J. W. J. Lowry in a passage now known as “Wombat Warren” extending north east from the chimney-like entrance. I visited the site with G. W. Kendrick and J. W. J. Lowry in 1969, collected further specimens, and made a cursory examination of the deposit. A steeply sloping talus cone descends into “Wom- bat Warren” from what appears to be an old entrance, now choked with sediment. Some specimens were found in or on this slope, but others were recovered from a poorly to well lithified red sandy deposit containing lumps of stalactite, ferruginous nodules, and charcoal as well as bone, suggesting it may be a lithified portion of an original talus cone below the postulated old entrance. This lithified deposit formed a ledge adhering to and projecting from the steeply sloping roof or wall of the chamber. A short note on material recovered was pub- lished (Merrilees 1969), but this did not include Petrogale. However, portion of the dentary of an adult macropod (specimen 69.4.2) mentioned in this report appears to represent Petrogale, and another juvenile Petrogale dentary (66.1.3) was found in 1965 near the cave entrance by B. G. Muir. No estimate of age is available for the Labyrinth specimens, but 69.4.2 was associated with Sthenurus and other extinct taxa in the lithified ledge. Specimen 66.1.3 has only slightly lithified matrix adhering to it (unlike 69.4.2), but the bone has a somewhat chalky appearance which often seems to indicate con- siderable age under south-western cave condi- tions. Deepdene Cave (AU1). — D. L. Cook collected Petrogale and other specimens from the site in Deepdene Cave about to be described and pre- sented them to the Western Australian Museum in 1958. He showed the site to P. Cook, who drew the attention of P. Henley to it, resulting in an excavation which produced a large Petro- gale sample from a small volume of deposit. This excavation was made by P. Henley and D. C. and J. W. J. Lowry and others in 1968, and described by D. C. and J. W. J. Lowry (1968). A radiocarbon date of 19 400 ± 1 200 yr B.P. (GaK — 2417, Kigoshi, Suzuki and Fukatsu 1973) was determined on the collagen of bone (mainly of juvenile Petrogale) from the lower half of the excavation. The cave is described briefly by Caffyn (1973). The Lowry and Henley collection was pre- sented to the Museum, and on analysis turned out to contain an overwhelming majority of fragments of very juvenile rock wallabies. For example in the lower half of the sample, there were 46 left upper milk molars of Petrogale, and hence at least 46 juvenile animals had con- tributed to the sample. There were probably no more than 2 adults of Petrogale, as judged on right first upper incisors with fully formed roots (there was only one permanent premolar with fully formed roots, the most certain indication of the presence of an adult). Minimum num- bers of individuals of other taxa were as follows: murids 3, Setonix 2, Bettongia penicillata 1, Potorous 4, Pseudocheirus 2, Trichosurus 3 and Isoodon 3; some of these were adult, some juvenile. Two tooth fragments in the deposit conceivably could represent Sarcophilus , but this is uncertain. The highly selected character of the sample was noted while the excavation was in progress (Lowry and Lowry 1968), and the excavators put forward three alternative suggestions as to its origin. The first was that the place might have served as the lair of some carnivore such as Sarcophilus. The second was that it might have been a human midden, and the third was that animals might have drowned while drinking at a pool in which the deposit was accumulating. To these suggestions may be added two others, that the pool did indeed serve as the lair or feeding place of a predator, but this was either a large owl or a large snake. In favour of the owl suggestion is the concentration of bone in one place, provided with ready-made perches and a platform on which prey could be deposited before being eaten. Against, perhaps, is the highly selected nature of the prey, and this favours the snake suggestion. Boids no larger than existing species might be expected to take young rock wallabies, Setonix, Trichosurus and possibly feeble or moribund individuals of even larger species, and there is a possibility that a boid very much larger than any of the existing Australian species was in- volved. Very large snake vertebrae are known from Mammoth Cave (Merrilees 1968) about 22 km north of Deepdene Cave, and from Koala Cave, near Perth (Archer 1973). These have been examined by Dr M. J. Smith, University of Adelaide, who believes them to be conspecific with Wonambi naracoortensis (pers. comm.). Wonambi may have reached a length of 5 m (Smith 1976) and the size of its vertebrae indi- cates a girth several times that of the living Australian pythons. I have not examined the site, but estimate from notes and diagrams supplied by the exca- vators that rather less than 10% of the deposit has been collected. A minimum number of 252 individual animals is represented, estimated by methods described by Baynes et al. (1976) as amended by Balme et al. (1978). Assuming that in this relatively large sample from a very circumscribed deposit the conventional “mini- mum number of individuals” approximates the actual number, then more than 2 500 individuals were taken by the predator concerned. If this was a large animal like Wonambi, it might have collected 1 prey animal every few days for part of each year, in which case some 20 years accumulation is indicated. The whole deposit might be attributable to one such animal. Strongs Cave (WI 63). — This is a very long tunnel-like cave traversed by a stream for its full length. A description and maps have been given by Williamson et al. (1977). It has a chimney-like entrance leading to a talus cone which has buried the stream, so that the water takes a concealed (but apparently very little 74 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. impeded) course through the base. From the point where it re-emerges for some 30 m down- stream, many of the small depressions in the stream bed contained isolated teeth of mammals, some of extinct taxa like Sthenurus (Cook 1963), among other “pebbles”. Extensive collections of these isolated teeth were made and presented to the Museum, including many of Petrogale e.g. the four molar enamel caps now included under catalogue entry 73.11.53. But these were not at first recognized as Petrogale, and in 73.11.53, for example, were labelled merely as “macropods”. The presence of the extinct taxa led to some detailed study of the cave by myself and others, and some systematic excavation was undertaken in what appeared to be a remnant of the ulti- mate source of the stream bed specimens. This was called “left ledge” in field notes, because it was on the left bank of the stream in the entrance chamber, and had a gently sloping surface standing several metres above the steeply sloping surface of the talus cone burying the stream. “Left ledge” and its opposite number “right ledge” appear to be the remnants of an older talus cone which has been undermined by the stream and let down in its central portion. This central portion may have received accre- tions of sediment, including teeth and bones, more recently than “left ledge”, and some of this younger material may have found its way into the stream bed. Thus the teeth collected in the stream bed might be of various ages. The excavation on “left ledge” reached what appeared to be an unfossiliferous basement of tumbled limestone blocks and fragments at very shallow depth. Nevertheless, it proved to con- tain extinct taxa (e.g. specimen 65.9.28, Sthenurus brownei ) and Petrogale (e.g. 65.9.11), suggesting some considerable age for at least some of the Petrogale specimens recovered. However, no radiometric dates are available for Strongs Cave. The older Strongs Cave fauna (ignoring extant taxa known only from surface litter or from the stream bed) is listed by Merrilees (1968). Although man, Sarcophilus and other preda- tory species are recorded from Strongs Cave, there is no other reason to postulate that it served as a predator’s feeding place. On the contrary, the nature of the cave entrance sug- gests a “pit trap” effect, accounting for the presence of the predatory as well as the non- predatory species. Devil's Lair (WI 61) . — There are many references to and descriptions of Devil’s Lair, four of which give analyses of mammal remains includ- ing Petrogale. These are by Lundelius (I960), Dortch and Merrilees (1972), Baynes et al. (1976) and — most significant for present purposes — Balme et al. (1978). The last paper re-examines suggestions made previously and attributes mammal (and other) remains in the deposit initially to owls (up to about 30 000 yr. B.P.) and then to human beings (sporadically), Sarcophilus (perhaps more consistently, between 77266 — ( 3 ) 75 short periods of human occupation) and owls (probably diminishingly after about 30 000 yr. B.P.) . Petrogale is first represented in the Devil’s Lair deposit in what Balme et al. (1978) desig- nate Layer 29 and estimate to be a little less than 30 000 years old. But its representation from this layer up to Layer 11 is very sparse, and includes some doubtful specimens such as upper molar 77.4.652 which is so worn as to obscure its Petrogale- like characteristics. How- ever, there are undoubted specimens in these older layers. Petrogale begins to be more abund- ant in Layer 10, and from Layers 6 to D (i.e. from somewhat later than 19 000 yr. B.P. up to the not precisely known but approximately mid Holocene time of formation of Layer D), it is fairly abundant. Flowstone layer D was partly buried by a black humic redistributed forest floor soil apparently about 300 years ago when the cave was re-opened in a new place. There was con- siderable disturbance of this uppermost black Layer A by the first excavators, and it is often difficult to decide whether a given segment has or has not been disturbed. Petrogale is recorded from Layer A, but I have re-examined the specimens concerned, and consider it possible that they are secondarily derived from excavated material. In view of the absence of historical records of Petrogale and of its absence from two dated deposits including late Holocene material — Skull Cave (Porter in press) and Deepdene Cliffs (Archer and Baynes 1973, a site near to but not identical with Deepdene Cave) — I reject the Layer A record in Devil’s Lair. Thus I infer that Petrogale vanished from the Devil’s Lair district, and perhaps from the Cape Leeuwin-Cape Naturaliste region, at some mid Holocene time not at present known more precisely, but sub- sequent to the formation of Layer D in Devil’s Lair. The relevant specimens are 73.10.61-63, 73.10.370-372, 77.6.358 (reported Layer A) and 73.10.399-402 (Layer D). It is possible that the first appearance of Petrogale in the Devil’s Lair deposit approxi- mately marks the time of its arrival in the region, i.e. rather less than 30 000 years ago. This is consistent with the record, first of sparse- ness and then of abundance, presumably as colonies became well established or the environ- mental trends which led to the immigration in the first place continued, or perhaps intensified. However, there is a proviso. The lowest parts of the Devil’s Lair deposit contain specimens some of which like Sthenurus, cannot be regarded as owl prey, which apparently pre-date occupa- tion of the cave by man or Sarcophilus, and many of which are more or less completely coated with a layer of cemented sand grains, and so present a characteristic appearance. Such specimens are regarded by Balme et al. (1978) as possibly secondarily derived (“re-worked”) from an older deposit which they postulate as occu- pying a position in or near the old entrance. The existence of such a deposit is not proven, and if it exists, its extent, nature and content Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. are known only from the small range of speci- mens which have become re-worked into the deposit sampled by excavation. Conceivably it could contain Petrogale. Indeed Petrogale is known to be associated with extinct taxa in The Labyrinth and Strongs Cave, and these associa- tions may well be older than 30 000 yr. B.P. Thus the Devil’s Lair indication of an arrival of Petrogale about 30 000 years ago and a local extinction about, say, 5 000 years ago can only be accepted tentatively. Giants Cave (WI 21 and 22) . — A single specimen, collected by P. J. Bridge in 1962, represents the Petrogale record for this cave. It is the horizontal ramus of a juvenile right dentary (65.12.51) which was partly encrusted with a granular matrix when found, suggesting it was not deposited very recently, but otherwise with no indication of its geological age. Giants Cave is not notable for any abundance of mammal remains, but it has yielded Petro- gale and Thylacinus (67.9.1) among locally extinct taxa. P. J. Bridge has collected Macro- pus fuliginosus post-cranial material (65.12.45) of chalky appearance cemented into a granular matrix, and I have collected fragments of an unidentified large macropod, also of chalky appearance, thickly coated with dense flowstone (66.7.3). It is possible that these specimens are of some considerable age. The cave consists essentially of a large tunnel opening at one end from a large doline with steeply sloping but not vertical walls, and by way of a talus slope from a small doline at the other end. It would not appear to have acted as a pit trap, so presumably the bones in it were taken there by predators, or represent animals which died in the cave. Museum Cave (WI 21). — A skull fragment labelled 12018 and several other fragments (in- cluding a right maxillary fragment) collectively recatalogued as 77.10.3 (originally stored with Setonix specimens) and a left maxillary and one other fragment catalogued as 77.10.1 (originally with Macropus irma specimens) appear to be parts of the same skull. If so, they would be part of a collection made in Museum Cave by L. Glauert (according to catalogue data with 12018 which are not in the collector’s hand- writing) and hence would have been collected in 1912, according to a newspaper article quoted by Mahoney and Ride (1975 p. 195). Despite this confusion, there is no reason to doubt the authenticity of the locality record, for an old and corroded label with the Macropus irma specimens mentioned above (66.9.63-70) carries a pencilled inscription which, though partly obscured, can hardly be other than “Museum Cave”; it appears to be in the collec- tor’s handwriting. There is a record by Glauert (1948), presum- ably the basis of one by Bridge and Shoosmith (1975), of Thylacinus from Museum Cave, but I can find no specimen to substantiate this record. Otherwise, the small collection of bone from the cave includes only extant species. The Petrogale skull fragments are fragile and partly encrusted, so that some considerable age may be postulated. A description of Museum Cave by Caffyn (1973) suggests to me that it may have acted as a pit trap for mammals. Yallingup Cave (YA 1). — This cave is in the north of the region, and although it has a long history as a tourist attraction and there are many published references to it (Bridge 1972), there appears to be no comprehensive descrip- tion of it as yet. However, one by Williamson, Loveday, Loveday and Bell is in preparation. My attention was drawn to it by the finding of a thylacine humerus and later a dentary (63.3.2, 63.3.42) during tourist development in 1963. I made a brief examination of this “thyla- cine locality” after this discovery. Later in 1963, as part of a tourist publicity campaign, D. Williams took up residence in the cave for several weeks. She made some shallow excavations in the cave floor in consultation with me, we corresponded, and I visited the cave again at an intermediate stage of these excavations. Although the work of an amateur, unused to interpreting and reporting excava- tions, they were made with care and attention to detail, and not under pressure of time. Taxa extant in the district in historic time (including Canis ) were recovered from several shallow excavations, but at the “thylacine locality” a deeper excavation was made, reach- ing what appeared to be a basal layer of jumbled fragments of limestone after traversing several distinct layers. The surface sloped quite steeply and the various layers sloped in consonance with it. The uppermost layer, containing the thylacine, was a well bedded sand, up to about 40 cm thick; apart from Thylacinus , mammals represented were of taxa still extant. This sand rested on a thin layer called “1st dripstone” in field notes and correspondence, meaning either a thin layer of sand rendered coherent by calcareous cement, or a thin crystalline flow- stone. Below this was a slightly lithified, rubbly layer containing bones, resting in turn on a “2nd dripstone”. The rubbly layer varied in thickness from about 60 cm to about 110 cm. Below the “2nd dripstone” was another slightly lithified, rubbly layer about 45 cm thick, resting on the basal layer of rocks. Excavation was discontinued at this rocky basal bed because excavated material began to be lost in open crevices between the rocks. No specimens were recorded from the “drip- stones”, but abundant mammal and some other remains were found in both layers below the respective dripstones. These layers consisted pre- dominantly of sand, but with numerous lime- stone clasts, the whole slightly lithified. They were excavated in arbitrary 7i cm, 15 cm or 30 cm “spits” because of their substantial thick- ness. Petrogale (63.7.184-189, 63.7.203, 63.8.9, 63.8.12-15, 63.8.19-22, 76.6.39, 76.2.99-100, 76.2.107-111, 77.10.2 and 78.1.66) is a conspicuous 76 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. component of the fauna recorded below “2nd dripstone”. Other taxa represented below “2nd dripstone” included the murids Notomys, Pseudomys albocinereus (neither known from the region in historic time) and Hydromys (indicating the presence of free water), together with an unidentified snake of medium size, Bettongia lesueur (not known in the district in historic time) and various mammal species still extant in the region. Snake vertebrae (76.2.112- 114) occur in all three of the arbitrary “spits” excavated below “2nd dripstone”; it is possible that these vertebrae derive from the same animal, emphasizing the arbitrary character of the divisions. Canis appeared just below “1st dripstone” (see Appendix 1). In another part of the same chamber in which D. Williams made her excavations, G. Pick also made excavations in an attempt to find exten- sions of the chamber. Excavated material was systematically sieved, bone was recovered, and its depth recorded. Towards the bottom of the tunnel so excavated, about 4 m below the cave floor, a single Petrogale tooth (77.8.51) was recovered. It is an upper permanent premolar, probably unerupted, though it appears to have undergone some root development. The enamel is somewhat mottled, there is adhering matrix, and this appearance and the depth of burial suggests that its age is considerable. The Petrogale samples Measurements All the available material consists of broken bones. I have studied only tooth-bearing speci- mens or isolated teeth, have measured only cheek teeth, and have concentrated statistical attention on the anterior cheek teeth (P 3 3 , dP\ P 4 4 and MS on the Thomas 1887 notation, or deciduous premolars, milk molars, permanent premolars and first molars respectively). Some data on posterior molars are included. All measurements were made by me in Novem- ber and December 1977 by applying the sharp points of dial vernier calipers over the tooth concerned in a plane judged to be perpendicular to the palate. There is some subjectivity in this, and although results were recorded as though correct to 0.1 mm, this probably exag- gerates their reproducibility; however, the prac- tical alternative of correcting to 1.0 mm would have underestimated their reproducibility. “Length” was measured in the plane of occlusal contact of the tooth concerned with those before and behind (or a projection of this plane in the case of the first and last teeth in a row). “Width” was a maximal record, for premolars, usually about central in a lower deciduous pre- molar, and posterior in a milk molar, a perman- ent premolar or an upper deciduous premolar. For a molar, “width” was used statistically only when the crown of the tooth concerned rose above the alveolar margin, as records for molars excavated from their crypts suggested that this measurement was otherwise unreliable. Anterior widths are recorded for molars, with posterior width added for M 4 4 , in which differences between anterior and posterior widths may be substantial. The Deepdene Cave sample was considered in three divisions for statistical purposes, the upper half and lower half of the Lowry and Henley excavation separately, plus a third division (“age uncertain”) made up of unlocalized specimens from this excavation and from the earlier Cook excavation. In recording measurements on Devil’s Lair specimens, three divisions, or grades of insight into relative ages, were recognized. The material covered by Balme et al. (1978), with numerous stratigraphic divisions and a series of radiocarbon dates, was regarded as most reliably dated. Next came material reported by Baynes et al. (1976) and Dortch and Merrilees (1972), based on thick stratigraphic divisions, and only tentative corre- lation between separate trenches. “Young”, “inter- mediate” and “old” specimens were separated in this division. “Young” means approximately early Holocene, and covers specimens recovered from Trench Al above the rubbly layer at 151-154 cm illustrated in Figure 2 of Dortch and Merrilees (1972), from Trench 6 above “brownish earthy layer” (Fig. 3 of Dortch and Merrilees 1973) and from Trenches 2 and 5 above “first orange brown earthy layer” (Figs. 4 and 5— left of Dortch and Merrilees 1973). “Intermediate” means late Pleistocene, (post-glacial-maximum) approximately, covering lower levels in Trench Al, “brownish earthy layer” in Trench 6, and “first orange brown earthy layer” in Trenches 2 and 5. “Old” means late Pleistocene, (pre-glacial- maximum) approximately and covers the lowest layers in Trench 5 and as far down as Layer 28 in Trench 2, but not including the contents of Pit 2 (see Fig. 4 of Balme et al. 1978). There were in fact very few specimens in the “old” group. Finally, an “age uncertain” division was recognized in Devil’s Lair, made up of specimens collected prior to the systematic series of excavations which began in 1970 plus those dislodged during extensive section cleaning or disturbed in other ways during the systematic excavations. For most statistical purposes, the accurately localized Balme et al. material was apportioned as “young”, “inter- mediate” or “old”. “Young” was taken to mean Hearth 2 and upward (excepting material allegedly from Layer A but here considered to be disturbed and hence in the age uncertain” division). “Intermediate” meant Layers 8 ( upward to M inclusive, with the contents of Pit 2. ‘ Old” meant Layers 9 and downward. The Yallingup Cave sample, although small, has been kept separate for statistical purposes, mainly because the northerly position of this cave might imply sub- stantial climatic differences between it and, say, Deep- dene Cave. The small samples from the other caves have been combined for statistical purposes. Two small samples of modern Petrogale from regions as close to the Cape Leeuwin-Cape Naturaliste region as were available, are also treated statistically. These are from Mondrain, Wilson and Combe Islands in the Recherche Archipelago, treated as a single sample (“Recherche”), and from a restricted area near Quaira- ding. Single specimens from other localities, mostly modern, were also measured; not only of Petrogale, but also of other macropods (Tables 9, 11). In the case of the posterior molars (M 2 2 , M 3 3 and M 4 4 ) the Deepdene Cave and Devil’s Lair samples have been treated as one, and arithmetic means for the whole fossil sample treated as one are also shown for all teeth measured, in Table 9. In each sample, measurements were made of teeth of one side only, nearly always the left. Thus each upper and each lower measurement of a given tooth represents a distinct individual, though that individual might be represented also by other teeth. Teeth so worn as to give a distorted impression of their original size were measured, but not included in statistical opera- tions. Isolated molar teeth, or sometimes even one or two teeth in a maxillary or dentary fragment, also were not included in the statistics unless their position in the tooth row could be established with confidence. Tables 1-11 summarize this metrical informa- tion. In them “N” means number of individuals in the sample, “O.R.” the observed range in size, “X” the arithmetic mean, “s” the standard 77 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. deviation, “V” the coefficient of variation, “t” the Student’s-t statistic for comparison of means. In these statistics and in the estimation of significance based on them I follow Simpson et al. (1960). Some characteristics o/ Petrogale teeth , and differentiation from other macropods Rock wallabies have teeth (Figs. 1-3) of typically macropodine form, as described and illustrated, for example, by Bartholomai (1975). But unlike Macropus fuliginosus, they retain functional premolars throughout life. (An exceptional modern specimen is M6208, which has lost its right upper permanent premolar, though all three other permanent premolars are still present and do not show exceptional wear). Molar progression often results in compression of the first molar against the permanent pre- molar, with consequent distortion of its shape and loss of its substance (e.g. 63.7.99c), or even in rare cases (e.g. M4094), total loss. M 2 2 also may be reduced in length (e.g. 77.5.558) or even width (e.g. 63.7.95c) by wear occasionally. Figure 1. — Petrogale. Occlusal view of 63.8.15 from Yallingup Cave, showing adult maxillary cheek teeth (P 1 , M 1_ 4). Figure 2. — Petrogale. Occlusal view of 77.5.473 from Devil’s Lair, in which P 3 resembles P 3 of Setonix. Other teeth shown are dP 1 and M 1 . Figure 3. — Right upper first incisors of very young rock wallabies from Deepdene Cave, lingual aspect. Tooth on right (part of 76.2.58) has accessory cuspule well developed, that on left (part of 76.2.59) has subdued buttress not terminating in a discrete cuspule. The molars are slow to erupt (as shown indirectly by the N columns of Tables 1-6), and the fourth molar appears never to erupt before the permanent premolar, by which time the first molar may be extremely worn (e.g. 77.5.638). In this respect, Petrogale differs markedly from Setonix , and is perhaps more extreme than the other two macropods ( Macro- pus irma and M. eugenii) with which it has been confused in the south-west. Lower molars apparently erupt before the corresponding uppers (e.g. M4426). Like many macropodines, but unlike the south-western potoroines, Petrogale shows pro- gressive increase in molar size from front to rear (Table 9), so that the fourth molars are much larger than the first (Tables 4 and 6). In this respect, Petrogale differs from Macropus irma, in which the gradient in molar size is generally less, and indeed in which the fourth molars may be smaller than the third (Table 9). The anterior portions of the molars in Petro- gale are generally larger than the posterior, and this is very noticeable (and is statistically sig- nificant — Tables 6, 7) in the fourth molars, to the extent that an isolated molar tooth, even if its size falls in the overlap in range between third and fourth molars (Tables 5, 6), may be identifiable as a fourth rather than a third by this posterior reduction. A given molar of Petrogale is generally smaller than the corresponding one in Macropus irma but larger than in M. eugenii and Setonix (Table 9) and if its position in the tooth row is known, this size difference is apparent to the naked eye, and confusion is unlikely. However, with molars of uncertain position, in which size 78 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. is a poor guide to identity, there are some dis- tinctions in form discussed by Merrilees and Porter (in press). For example, the buccal portion of the anterior shelf in M 3 ,4 is more completely enclosed by a low marginal rim in Petrogale than in Macropus irma, while the anterior shelf as a whole is relatively narrower and more “nose-like” in all lower molars of M. eugenii than in Petrogale. In M 2 , 3 , 4 in Petro- gale, unlike M. irma or M. eugenii, the median valley is closed by a longitudinal low crest or partial cingulum on its lingual side. The molars of Setonix are distinctive and small, and hence not likely to be confused with those of Petrogale. On the other hand, both deciduous and per- manent premolars are large relative to the molars in Setonix (Table 9), and are comparable in size and in shape with those of Petrogale, so that the two taxa are easily confused, whereas these premolars are small relative to the molars in M. irma and M. eugenii and differ markedly in shape from those of Petrogale. But care is needed to distinguish an isolated upper perman- ent premolar of M. eugenii from an isolated upper deciduous premolar of Petrogale. There is also some possibility of confusion in milk molars in that the upper milk molar in M. irma, though larger than that of Petrogale, is not unlike it in shape. Some criteria for distinguishing premolars and milk molars of Petrogale from other taxa with which confusion is possible are as follows: — 1. Deciduous premolars. In Petrogale, P 3 is likely to be marginally longer and narrower than in Setonix (Table 9) giving a visual impression of narrowness in direct comparisons. In both taxa there is a central longitudinal depression, which is subdivided by more numerous and better developed (but still minor) transverse crests in Setonix. The most anterior and most posterior of the small compartments so formed in this depression are narrower and more obviously inclined to the longitudinal axis of the tooth in Petrogale. In M. eugenii, P 4 is usually shorter and narrower than P 3 in Petrogale (Table 9), and the lingual low longitudinal crest which in P 3 of Petrogale (and Setonix) helps to define a central depression is confined to the rear part of the tooth in P 4 of M. eugenii, so that there is no anterior compartment. In Setonix, P ;{ is generally marginally shorter and wider than in Petrogale, giving a visual impression of its being a stouter tooth, which impression is strengthened by its having its maximum width anterior, whereas Pa in Petrogale is quite narrow anteriorly. 2. Milk molars. As in most macropodines, the milk molars in Petrogale are generally molariform in shape, and yet easily recognizable as milk molars (dP 4 4 ) by their anterior constriction and distinctive anterior shelves. The most likely confusion would appear to be between Petrogale and M. irma in dP 4 ; however the latter is substantially larger (Table 9). Further in d _f 4 in Petrogale, the buccal corner of the anterior shelf is more angular in plan, the fore- and mid- links more pronounced, and the median valley more definitely truncated buccally by fusion of crests ascending from the paracone and metacone, than in M. irma. 3. Permanent premolars. In Setonix, P 4 appears to be rather variable in size and shape, but is often somewhat longer, lower crowned and more uniform in width than that of Petrogale, which is considerably wider posteriorly than anteriorly (Fig. 1). Likewise P 4 in Setonix tends to be longer and wider than in Petrogale, but there is a good deal of overlap (eg in Table 9) and in both taxa, P 4 is blade-like, with a longitudinal crest which is inflected lingually at the rear. In relatively unworn teeth, this crest is seen to 79 be slightly serrated, because of a longitudinal succes- sion of cusps. The most anterior of these tends to be the most prominent in Petrogale , whereas in Setonix the two most anterior are more nearly equal in prominence. There is a longitudinal cingulum, not very pronounced, but sometimes present on both buccal and lingual faces of P 4 in Petrogale, whereas it is usually very inconspicuous or missing entirely from the buccal side in Setonix. In both taxa, a crest descends from the most anterior cusp to form the front of the tooth, and in Setonix this front crest projects further forward at the enamel margin than in Petrogale, in which the descent is more nearly vertical. But none of these differences is marked, and in practice it is sometimes very difficult to decide on the identity of an isolated P 4 . Fortunately, where any of the more posterior teeth is present, the distinc- tion between Petrogale and Setonix is clear. Lower incisors in Petrogale and Setonix are similar in shape and size, and retain a leaf-like appearance for a larger proportion of the ani- mal’s life than in M. irma or M. eugenii. They are generally slightly narrower (laterally) where they emerge from the bone in Petrogale than in Setonix, and are more incumbent than in Setonix, M. irma or M. eugenii. Upper incisors in all four taxa are similar in general form, i.e. the first is a slightly curved blade, the second of approximately equidimen- sional section, and the third lobate. Those of M. irma are noticeably larger than of Petrogale, Setonix or M. eugenii. In the last three taxa, size differences are a poor guide to identity, and are often obscured by wear differences. A small peg-like cuspule standing out from the lingual face of I 1 , originating about half way down in an unworn tooth, nearer to the posterior border, is very characteristic of Petrogale where present (Fig. 3). However, as discussed under “vari- ability”, this cuspule is not always present, and a somewhat similar cuspule occasionally appears on I 1 in other taxa (e.g. 77.6.142 or M1760, M. irma) . Detailed criteria for distinguishing uoper incisors are given by Merrilees and Porter (in press). variability in Petrogale teeth The coefficients of variation in Tables 1-6 show not only that populations in a given place and time are fairly or very uniform in tooth dimensions (note Tables 2, 4 especially), but also that they are still uniform when time is dis- regarded (Deepdene Cave and Devil’s Lair entries in Tables 5, 6). However, somewhat greater variability in deciduous and (more notably) permanent premolars, and in widths rather than lengths, is recorded in Tables 1 and 3. Time can be taken into account most clearly in the succession summarized in Table 10. So fai as statistical tests on the small stratigraphic samples of Table 10 have meaning, it may be noted that the difference between the means of length of P ! in Layer Q and Layers 4 5 6 combined is significant (5% level t = 260 degrees of freedom = 10), as also for length’ of P* m Layer O and Layers Y and Z combined (t — 3.80, degrees of freedom = 8). These and some other differences which are not statis- tically significant (e.g. length of dP 4 in Layer O compared with Layers 10 d and e) suggest there was some tendency for tooth size to increase in the Petrogale population around Devil’s Lair as Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. time went on. But other differences (e.g. between length of M, in Layer O compared with Layer 6 — not statistically significant) are opposite in trend, and none of the apparent trends is very marked. There are some variations in shape as well as size among Petrogale premolars, and many of these variants resemble Setonix even more closely than modal specimens do. A fairly common variant of P :! in Petrogale is the development of a “pocket” in the enamel of the buccal face towards the front of the tooth (e.g. 77.5.473 Fig. 2). Sometimes this is accompanied by an enamel fold towards the back of the tooth (e.g. 75.4.243). Particularly in the latter case, the tooth comes to resemble a P 3 of Setonix quite closely. In fact, 75.4.243, an isolated tooth, was initially identified as Setonix, and even after re-examination and ascription to Petrogale, remains a somewhat doubtful case. Partial development of a buccal cingulum, as these structures in P ! perhaps could be described, may occur also in P ;i . For example, in 73.11.807, the middle third of the buccal face of the tooth carries a distinct cingulum, developed into a deep pocket in the enamel of the posterior third. Another variant of P ;J is extreme lingual inflexion of the posterior part of the crest forming the blade-like occlusal part of the tooth. In 77.6.483, this posterior portion of the crest is transverse, approximately perpendicular to the (major) anterior portion of the crest, which is nearly parallel to the longitudinal axis of the tooth. Variants of the modal milk molar form appear not to be common, a matter of some importance in the discussion (in Appendix 2) of 70.12.1132. However, in 77.12.1, the forelink is interrupted, the more anterior portion crossing the anterior shelf obliquely, whereas usually the forelink rises continuously from the anterior shelf to the protoconid. As noted previously, P' in Petrogale sometimes has a longitudinal cingulum near the base of the buccal face. This is very subdued, barely noticeable in most cases, but in a few (e.g. 77.5.631) expands posteriorly into an enamel fold or even small “pocket” which could be described as an accessory cuspule. Similarly, a subdued cingulum may be present along part of the buccal face of Pi (e.g. 10744), or a small crest or enamel fold may extend slightly obliquely up the rear of the buccal face to fuse with the main central crest (e.g. 77.5.638). In 73.8.352 (an isolated P4, possibly never erupted) the central portion of the main central crest of the tooth descends far below the front and rear portions, but this appears to be a malformation rather than a functional variant. The small peg-like cuspule on the lingual face of I 1 in Petrogale, described in the previous section, is present in a large proportion of teeth, indeed in some samples in a majority (e.g. 37 out of 63 left and 46 out of 86 right upper first incisor teeth in the “Deepdene Cave — upper” sample). In most samples, however, a discrete peg is present only in a minority of teeth (e.g. “Deepdene Cave — lower”, in 19/42 right and 15/38 left I'), though in virtually every case there is some corresponding structure ranging from a slight swelling in the enamel to a well marked vertical buttress. In 77.11.50, this but- tress culminates in a small and a larger projecting peg, and all or most of the single pegs similarly represent the culmination of buttresses. The distribution of this peg-like cuspule appears to be random both in space and time. Thus in Devil’s Lair, incisors with and without it occur in the same layer, often in the same trench, e.g. in Trench 7d, Layer Y contains 75.4.444 (with cuspule) and a specimen without cuspule stored with 77.5.532. There are teeth with these cuspules in most layers in Devil’s Lair from at least as high in the sequence as Layer M (73.10.204) to as low as Layer 28 (76.9.225), almost the lowest layer containing Petrogale. There are numerous instances in both Deepdene Cave and Devil’s Lair, though none in the small samples from the other fossil sites discussed, nor in the modern samples used for statistical comparisons. However, it does occur in modern specimen M9872 from the Warburton Range, some 1 400 km from Deepdene Cave, in what appears to be the same species as Devil’s Lair and Deepdene Cave. These morphological variants appeared to be randomly distributed both in time and space, and I was unable to identify any constant or progressive differences in tooth form. Discussion and conclusions Specific identity of the south-western fossil rock wallabies Calaby (1971) suggests that the highly dis- continuous distribution of rock wallaby popula- tions has resulted in the appearance of numerous size and colour variants. Many of these have been distinguished as “species”. Both Ride (1970) and Calaby (1971) recognize more than one species in northern Australia, and Kitchener and Sanson (1978) have recognized an additional species recently. But there appear to be only two species in southern Australia, Petrogale xanthopus and the very wide ranging and vari- able P. penicillata, taken by Ride (1970) to include “P. lateralis ”, “P. liacketti” and “P. pearsoni”, the named south-western “species” (see Serventy 1953 for authors of and comments on the status of these taxa). The south-western fossil samples can be inter- preted as supporting the concept that discon- tinuous populations with distinctive characters can be regarded as conspecific. It is highly probable that the Devil’s Lair sample represents a single population persisting for some 25 000 years (Balme et al. 1978) with only minor changes in tooth size (Table 10 and comments in the preceding section), and rather uniform tooth sizes over any given part of this time (Tables 1-6). With a lower but still high degree of probability, it may be sug- gested that the Deeodene Cave and “other caves” Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. samples represent populations conspecific with that round Devil’s Lair, and (at a slightly lower level of probability) with that around Yallingup Cave. Yet there are numerous statistically sig- nificant differences in tooth size (Table 7) in these samples. Thus it would seem reasonable to give little biological weight to the statistically significant differences between south-western fossil samples and the modern Recherche Archipelago samples in some dimensions, nor between the Recherche and Quairading samples in length of P 4 (Table 7). Nor do there appear to be any constant differences in tooth form among any of these samples. Serventy (1953) reports an opinion that Wilson Island might have a population of smaller rock wallabies than Mondrain, Combe and Salisbury Islands, in the Recherche Archipelago. Jones (1924) suggests that the Pearson Island animals (South Australia) might be smaller than those on the mainland. But if these observations are accurate for general body size, it does not seem that tooth size is necessarily closely related to body size according to the comparison of single specimens shown in Table 11. I have not made the very extensive metrical and morphological studies of modern samples of Petrogale penicillata and other rock wallaby species which would be required fully to justify inclusion of the south-western fossil samples under P. penicillata. But for practical purposes, in the absence of any strong evidence to the contrary, I assume they can be so included. Arrival and extinction of rock wallabies in the south-west Except for a mislabelled modern specimen discussed by Baynes (in Baynes et al. 1976 p. 125), there are no rock wallaby remains in a macropod-rich deposit containing Sthenurus, Zygomaturus and numerous other extinct taxa in Mammoth Cave, with an age in excess of 37 000 yr B. P. (Merrilees 1968). But the undated deposit in Strongs Cave which contained Petro- gale also contained Sthenurus, and the Labyrinth has yielded both Petrogale and various extinct taxa, probably in association (see notes on deposits, above). In the lowest parts of the Devil’s Lair deposit, rock wallabies are absent until about 30 000 yr. B. P., cannot be described as abundant until about 21 000 yr. B. P., are still present up to Layer D, estimated to date from about 5 000 yr. B. P., but are not certainly present in undis- turbed parts of the topmost layer dated at about 300 yr. B. P. At about 19 000 yr. B. P. there was a flourishing colony near Deepdene Cave. No rock wallabies occur in a deposit in Skull Cave, only 5 km from Deepdene Cave, over a time range covering most if not all the Holocene (Porter in press), nor are there any records of them in the Cape Leeuwin-Cape Naturaliste region in historic time (Baynes et al. 1976). It would appear that rock wallabies may have arrived in the Cape Leeuwin-Cape Naturaliste region at some tim e prior to 30 000 yr. B. P., though they might not have built up substantial populations until nearly 20 000 yr. B. P. Balme et al. (1978) suggest that their time of arrival in the Devil’s Lair district approximates the time of their first appearance in the deposit. But this suggestion may not be consistent with the Strongs Cave evidence (and possibly that from The Labyrinth) of contemporaneity of Petrogale with various large extinct taxa. At any rate, rock wallabies seem to have arrived between the unknown time of formation of the Mammoth Cave deposit and 30 000 yr. B. P., and to have disappeared from the region well before historic time and perhaps not long after the time of their last appearance in the Devil’s Lair deposit, i.e. after 5 000 yr. B. P. Implications of the rock wallaby migration and extinction Such a record of invasion and extinction raises interesting questions: by what route, and under what kind of climate and vegetation was the invasion possible, and what changes brought about the extinction? Since rock wallabies were not present under the vegetational and climatic regime of historic time, this was presumably inimical to them. For the geologically short period under discussion, it is reasonable to assume that if rocky outcrops favourable to rock wallabies once existed, they still exist, and to seek evidence of environmental changes other than in habitat. There appears to be little doubt that one such environmental difference was vegetational. With- out postulating any climatic difference, one must envisage a much greater extent of coastal plain in late Pleistocene than in late Holocene time because of glacio-eustatic effects. By postulat- ing additionally a cooler, drier and windier climate during and for some thousands of years subsequent to the time of minimum sea level, one sees this wider coastal plain as a wider belt of heath or scrub formations than exists at present. If wind was, as it still appears to be, a major determinant of vegetational boundaries, it may be that this wider late Pleistocene coastal heath or scrub adjoined forest or other tree formations more or less along the same boundary as at present, greater wind speeds being counter- balanced by greater distance from the sea. The likely effect of reduced effective rainfall in the late Pleistocene presumably would be reduced plant cover in any formation, but on present evidence it is difficult to estimate the extent of this reduction. Balme et al. (1978) present evidence from changes in the mammal fauna in Devil’s Lair over a period beginning about 35 000 years ago and ending about 5 000 years ago, but point out that mammals may be less sensitive to climatic fluctuations, and their remains less reliable indices of such fluctuations than practically any other animal group and certainly less reliable than plant fossils, ’con- sequently, their interpretation of the changes reflected in the Devil’s Lair mammal fauna provides for a wide range of possibilities. Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. At one extreme, these faunal changes could be consistent with a climate only a little drier than at present, at the other extreme with a climate so much drier that profound vegeta- tional differences from the present must be envisaged. One possibility is that about 35 000 years ago some plant formation sufficiently open to support a substantial population of Tarsipes was a major one near Devil’s Lair, and that it remained so for several thousand years. It may have given way to woodland or forest by 20 000 years ago, but this may have been a jarrah- marri association unlike the karri high open forest of historic time. Such karri forest may have come to dominate the southern part of the Cape Leeuwin-Cape Naturaliste region only after 10 000 years ago. The northern part of the region, including Yallingup Cave, at present is vegetationally different from the southern (Smith 1973), and it is perhaps reasonable to postulate that if Devil’s Lair was surrounded at any stage by, say, jarrah-marri high woodland, Yallingup Cave would have been surrounded by jarrah- marri woodland or banksia low woodland or some other more open formation. So far as rock wallabies are concerned, if it can be assumed that karri high open forest or jarrah-marri open forest (i.e. the main forma- tions occupying in historic time those localities known to have harboured rock wallabies in prehistoric time) are inimical to them, then it can be assumed that other more open formations were dominant when rock wallabies invaded and flourished in the district. Wakefield (1971) suggests that a change from grassy parkland to dense tough shrubbery in western Victoria in historic time was unfavourable to Petrogale penicillata, lending support to the suggestion that density of plant cover is in some way influential. If a present general characteristic of P. peni- cillata is a very wide geographical range occupied by discontinuous, widely separated populations, it is reasonable to suggest either that the species has an impressive ability to colonize and spread across unfavourable environments until it reaches favourable ones or that the present isolated populations are relics from an origin- ally continuous one. (It would be ironic if “rock wallaby” really meant “wallaby preserved by rocks in an otherwise unsuitable environ- ment” rather than “wallaby specialized for life among rocks” as usually understood.) In the case of Petrogale penicillata migrating into the Cape Leeuwin-Cape Naturaliste region in late Pleistocene but pre-glacial-maximum time, a high degree of colonizing enterprise would appear to have been necessary. Not only would they have had very little choice of rock type, but they would have had to cross large areas completely devoid of rock outcrop of any kind. The geology of the region is discussed by Lowry (1967), and if due allowance be made for a zone now submerged and not well known geologically, but emergent during the late Pleistocene, Lowry’s map covers all the possible migratory routes to the narrow belt of dune limestone which contains all the Petrogale sites described above. Perhaps the most likely would be from the dissected west-facing scarp of the Darling Fault to the Whicher Range and thence to the elevated and dissected Leeuwin-Naturaliste Block. But the Whicher “Range” is a very subdued chain of low hills, quite unlike the rock wallaby habitats in Victoria illustrated by Wake- field (1963, 1971). Another more remote possibility would be along the banks of the Blackwood River if that were much more deeply entrenched than it is now, during a period of lower sea level. The Leeuwin-Naturaliste Block is sufficiently dissected to be credible rock wallaby territory. There are collapse dolines round many caves in the central “spine” of dune limestone, occasional outcrops of gneissic basement rock, and sea cliffs in both kinds of rock. There may have been cliffs along earlier western coastlines, now submerged and blanketed by sediment. But it seems very unlikely that earlier coastlines of Geographe Bay, more or less parallel to the present coast in the vicinity of Busselton, but north of it, would have been cliffy, and hence potential migration routes for Petrogale. If in addition to these geologically unfavour- able conditions, the immigrant rock wallabies had to contend with unfavourable (i.e. dense) plant formations, it would seem unlikely that they would have reached the haven of the Leeuwin-Naturaliste Block. Therefore it is reasonable to suggest that plant formations were more open than at present. Of the range of possibilities allowed by the evidence of Balme et al. (1978), conditions close to the driest extreme may be selected as most conducive to rock wallaby migration. With colonies once established in the Leeuwin- Naturaliste Block, no doubt there could be changes in climate and vegetation, at least in some directions, which would still permit survival. But the eventual local extinction of these rock wallaby colonies and other “non- forest mammals” could be a function of climate. Replacement of Pleistocene open plant forma- tions by karri high open forest in the south of the region, and jarrah-marri open forest in the north and east would appear to be consistent with the rock wallaby record as well as with the known vegetational record in historic time and with the Holocene record studied by Churchill (1968). Differences in the south-western rock wallaby samples The analysis of the Devil’s Lair fauna by Balme et al. (1978), on which this review of south-western rock wallabies largely depends, is itself dependent on accurate identification of fragmentary remains. These authors suggest that closer analysis, taxon by taxon, should provide some quantitative estimate of their degree of accuracy. In this instance, at least 800 specimens identified by them as Petrogale were re-examined by me in a specialized context later. In my opinion, less than 40 had been 82 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. misidentified. It seems unlikely that this order of inaccuracy would influence their findings appreciably. As shown in Tables 1-6, the Deepdene Cave rock wallabies had generally longer, sometimes wider and sometimes narrower cheek teeth than those around Devil’s Lair, and these in turn had larger teeth than the Yallingup Cave animals. Some of these differences are statistically signi- ficant, some not (Tables 7 and 8, in which the sample with the larger mean is entered first in each pair). In the cases of significant difference the Deepdene Cave mean values show up as the largest. The “other caves” sample, drawn from the same population as that of Devil’s Lair, or from populations not far to the north, does not appear to differ markedly from the Devil’s Lair sample. Thus it would appear that some factor, presumably environmental, favoured develop- ment of larger teeth in the extreme south-west corner of the prehistoric rock wallaby range. As shown above, tooth size may not be closely related to general body size, so it is difficult to suggest what these favourable environmental factors were. However, there are pointers in Table 11 to a suggestion that equable climate (in the sense of Axelrod 1976), cool summers, or some other climatic factor coupled with main- land (as against island) ranges, are involved. Table 11 gives data on single specimens taken from the collection readily available to me and covering a large proportion of the western part of the range of Petrogale penicillata, mostly modern, but including prehistoric occurrences from caves north of Perth. In so far as single specimens can suggest trends, and taking Table 11 in conjunction with Tables 1-6, it would appear that island and inland specimens have smaller teeth than coastal, especially far south western, specimens. It may be that the differences between the Deepdene Cave and Yallingup Cave populations were a regional expression of factors operating over half the continent, but much more detailed studies than mine would be needed to identify these factors. At the one site where a long temporal succes- sion can be examined, namely Devil’s Lair, the samples are very small from layer to layer and trends correspondingly difficult to discern. From about 20 000 to 12 000 yr B. P., there may have been a slight increase in premolar size and a slight decrease in molar size in the rock wallabies. But if so, the differences were small compared with those involving different localities even in the circumscribed Cape Leeuwin-Cape Naturaliste region. In this region, there is a chain of sites, a few kilometres apart in the north, and even more closely spaced in the south, some of which are known to have harboured rock wallabies, and many of which might have done so. This is not a geographical situation in which one would expect to find much variation in the wallabies, yet they did vary in tooth size. I conclude that rock wallabies are sensitive to small climatic, vegetational or other differences in their immediate environment, and establish appropriate tooth sizes and perhaps other characters readily. But once established these characters remain fairly constant for long periods. Such a view is consistent with that taken by Calaby (1971) and others of the present condition of Petrogale penicillata, accounting for the taxonomic uncertainities surrounding this species. Acknowledgements. — I am grateful to the speleolo- gists mentioned in the text for specimens and descriptions of caves and cave deposits, to the Aus- tralian Research Grants Committee for funds enabling much of the basic studies of Devil’s Lair to go forward, to J. K. Porter, J. Balme, D. J. Kitchener and A. E. Cockbain for constructive criticism, to M. Wallis for typing the manuscript, and to J. K. Porter for the photographs here reproduced as Figures 1-3. References Archer, M. (1973, nominally 1972). — Phascolarctos (Marsupialia, Vombatoidea) and an associ- ated fossil fauna from Koala Cave near Yanchep, Western Australia. Helictite, 10: 49-59. Archer, M. (1975, nominally 1974). — New information about the Quaternary distribution of the thylacine (Marsupialia, Thylacinidae) in Australia. Journal of the Royal Society of Western Australia, 57: 43-50. Archer, M. and Baynes, A. (1973, nominally 1972). — Prehistoric mammal faunas from two small caves in the extreme south-west of Western Australia. Journal of the Royal Society of Western Australia, 55: 80-89. Axelrod, D. I. (1967). — Quaternary extinctions of large mammals. University of California Publi- cations in Geological Sciences, 74. Balme, Jane, Merrilees, D. and Porter, J. K. (1978). — Late Quaternary mammal remains, spanning about 30 000 years, from excavations in Devil’s Lair, Western Australia. Journal of the Royal Society of Western Australia 61: 33-65. Bartholomai, A. (1975). — The genus Macropus Shaw (Marsupialia: Macropodidae) in the Upper Cainozoic deposits of Queensland. Memoirs of the Queensland Museum, 17 : 195-235. Baynes, A., Merrilees, D. and Porter, Jennifer K. (1976, nominally 1975). — Mammal remains from the upper levels of a late Pleistocene deposit in Devil’s Lair, Western Australia. Journal of the Royal Society of Western Australia, 58: 97-126. Bridge, P. J. (1972). — W.A. cave list no. 6. The Western Caver, 12: 179-182. Bridge, P. J. (1973). — W.A. cave nomenclature list no. 11 The Western Caver, 13: 149-155. Bridge, P. J. and Shoosmith, R. K. (1975). — W.A. cave nomenclature list no. 13. The Western Caver, 14: 149-163. Caffyn, P. (1973). — Deepdene Cave AU 1 description and Museum Cave WI 31 description. The Western Caver, 13: 229-230. Calaby, J. H. (1971). — The current status of Australian Macropodidae. The Australian Zoologist, 16: 17-29. Cook, D. L. (1963). — The fossil vertebrate fauna of Strongs Cave, Boranup, Western Australia. The Western Australian Naturalist, 8: 153-162. Churchill, D. M. (1968). — The distribution and pre- history of Eucalyptus diversicolor F. Muell, E. marginata Donn ex SM. and E. calophylla R. Br. in relation to rainfall. Australian Journal of Botany, 16: 125-151. 83 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. Dortch, C. E. and Merrilees, D. (1972, nominally 1971). — A salvage excavation in Devil’s Lair, Western Australia. Journal of the Royal Society of Western Australia, 54: 103-113. Glauert, L. (1948). — The cave fossils of the south-west. The Western Australian Naturalist, 1: 100-104. Jones, F. Wood (1924). — “The mammals of South Aus- tralia” Pt. 2. The bandicoots and the herbi- vorous marsupials. Govt. Print., Adelaide. Kigoshi, K., Suzuki, N. and Fukatsu, H. (1973). — Gakushuin natural radiocarbon measure- ments VIII. Radiocarbon, 15: 42-67. Kitchener, D. J. and Sanson, G. (1978). — Petrogale burbidgei (Marsupialia, Macropodidae) a new Rock Wallaby from Kimberley, Western Australia. Records of the Western Austra- lian Museum, 6: 269-285. Lowry, D. C. (1967). — Explanatory Notes, Busselton and Augusta, Western Australia. Geological Survey of Western Australia 1:250 000 Geo- logical Series. Lowry, D. C. and Bain, T. D. (1968). — The Labyrinth, Augusta. The Western Caver, 8: 17-18. Lowry, D. C. and Lowry, J. W. J. (1968). — Hasty notes on the Deepdene Cave bone deposit. The Western Caver, 8: 32-33. Lundelius, E. L. (1960). — Post Pleistocene faunal suc- cession in Western Australia and its climatic interpretation. Report of the International Geological Congress, XXI Session, Norden 1960 Part IV: 142-153. Mahoney, J. A. and Ride, W. D. L. (1975). — Index to the genera and species of Fossil Mammalia described from Australia and New Guinea between 1838 and 1968. Western Australian Museum Special Publication 6, Perth. Merrilees, D. (1968). — Man the destroyer: late Quater- nary changes in the Australian marsupial fauna. Journal of the Royal Society of Western Australia. 51: 1-24. Merrilees, D. (1969). — A newly discovered bone-bearing deposit in Labyrinth Cave, near Augusta, Western Australia. The Western Australian Naturalist, 11: 86-87. Merrilees, D. and Porter, J. K. (in press). — Guide to the identification of teeth and some bones of native land mammals occurring in the extreme south west of Western Australia. Field Guide Series, Western Australian Museum. Milham, P. and Thompson, P. (1976). — Relative anti- quity of human occupation and extinct fauna at Madura Cave, southeastern Western Australia. Mankind, 10: 175-180. Porter, J. K. (in press). — Vetebrate remains from a strati- fied deposit in Skull Cave, Western Austra- lia, and a review of their significance. Journal of the Royal Society of Western Australia. Ride, W. D. L. (1970). — A guide to the native mammals of Australia. Oxford University Press, Melbourne. Serventy, V. N. (1953). — Mammals. Part 4 of “The Archipelago of the Recherche”. Australian Geographical Society Reports, 1 (4) : 40-48. Simpson, G. G., Roe, A. and Lewontin, R. C. (1960). — Quantitative Zoology. Harcourt, Brace, New York, revised edition. Smith, F. G. (1973). — Vegetation map of Busselton and Augusta. Western Australian Department of Agriculture. Smith, M. J. (1976). — Small fossil vertebrates from Victoria Cave, Naracoorte, South Australia. IV. Reptiles. Transactions of the Royal Society of South Australia, 100: 39-51. Thomas, O. (1887). — On the homologies and succession of the teeth in the Dasyuridae, with an attempt to trace the history of the evolution of mammalian teeth in general. Philo- sophical Transactions of the Royal Society of London, (B) 178: 443-462. Wakefield, N. A. (1963). — Notes on rock-wallabies. The Victorian Naturalist, 80: 169-1.76. Wakefield, N. A. (1971).— The Brush-tailed Rock- wallaby (Petrogale penicillata) in western Victoria. The Victorian Naturalist, 88: 92-102. Williamson, K., Loveday, B. and Loveday, F. (1977, nominally 1976). — Strongs Cave and related features — southern Witchcliffe, W.A. The Western Caver, 16: 48-63. 84 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. a oo o J-( vo vo or CM vo oT vo X s mm mm r-H h r-H t-K t-H vH r-H * n d< 666 m co or 66 6 oT CO oT t—H CO 6 or Xi CM CM (N CM CM CM CM 6 6 -a £ vo vo oT VO to co vo pi B M n to Tt OT r-H „ e fN OfS CM r-H M B 666 6 6 6 6 M cu X mm r- 1*** t- VO VO f- •o •o to JC 00 or or or or or or or or a> On 00 ON ON o o to to hJ O.R. mm TTT tSion or to o i i i or co oT or 4P o t or or or or oT oT or aJ co or or ■'t CO or vo s mm o or to to or or or or to or £ to vo CM m CM CM «o O CO co CM CO or o o a S a m eepdene Cave- Upper Lower Age Uncertain evil’s Lair — Young Intermediate Old Age Uncertain allingup Cave- Age Uncertain ther Caves — Age Uncertain odern — Recherche Quairading P P O * 85 Any position, usually central. Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. n CJ -O cl H to S*. .3 5 § > >o co (N co co NO CO CO co (N c CO CN ai B *0 Tf" 7 6 B NO OO Tf NO NO I-* r- Tf rt z no «o to COTfOh co co co O nO co i/N (N Tt «o fS CL g co co co Tt NO Tj- •O co c — (N — (N (N fN co ( C/5 c , E O O 6 o o o 6 6 cl ^Ix E i £ E NO NO NT) 04 Tf o NO »0 to to •O •o oo C 0 «o »o z O Tt Tt MiOCOTt CN CO tT O NO Tf • cs , : : c3 > '3 1 03 03 o3 • £ • ct3 • S-. to CO CJ a> v -' o ■- *5 4) o 0 JS.s i h O 86 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. 3 C3 H > VO 00 Ov VO rf 00 pi £ (N fS Tt m CN rm vo ~ co r- rf r- Tf ro m pi E VO VO VO vo VO VO VO VO o £ rN vo ov 00 © VO VO vo *0 u~) VO •o o z 'ton n«o-st^ (N rm oo ^ a to o > go r- oo cm VO 00 Tf S-i PH CU - 1 mm Tf T (N rm sf -C £ © © © © © © © o wid X mm oo © oo m fO Tt (N oo m 4-0 m m m rm t-i 5*. o u. — to r- av >0) ro m rm r" ro LULU S m (N • <33 : 03 1 03 as aJ • • • o3 ■ i- C3 IZ 1 'u. M C/5 CJ « ^ o b •- 0) •2 *U O u n ft r* -c .5 (U c c3 c o *o S3 i) (o — ' a^o m J c^ ^ ©: o *;2 oo 3 D oo •E a cd — > u S) i|.g ^■5 «« •rt O 3 13< £< oQ o3 00 o a> PL. > o >o CO IO £ \X £ co co co ro co CO CO co CO CO U _o »0 OS r- os r- vo vo r- L d pP E m m m m m m c ^ < ° £ m m m i-H m ro o co co CO CO CO CO CO CO £ t-( SO — ( Tt oo O O CO Tt 1— 1 — 1 (N CN > CN (N (N (N so CO T— < co C /2 £ _ < — < CO co nnN ’Cj- so CO SO G CN o o o 00 o IO SO 7 co oo so so r- o cn (N r- : : 1 : V : CS > ^ 1 a 3 aJ : tP • L 03 XZ 1 tp C /2 u 8 v!- ^ o o r_) o w o cn a> a> o x: •£ (U c 03 id c a c > c 1 H "2 (1) — ' X> D. £ oo -a w 3 £><£ o< •£< ■£< oc^a 1 Q Q O 88 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. > •o tj- vo > vo rT • 04 ol •o > >0 CO •rs c — 04 — co p ^r 04 co c/3 £ 66 6 6 C/3 £ 6 6 6 C-I CO IX £ Tf — 04 «0 VO S IX £ OOXh O' — -C £ VO VO VO VO VO VO £ O' O' vo VO vo vo 6X) 00 c <0 Tf CO > •o N s c/) £ ro 04 04 CO s C/3 £ 04 co V £ 6 6 6 E 6 6 X! JU 3 C3 H <3> 5 ■^s -o IX mm vo o- n- ^t- o «o >o 5-1 50 12 £ lx mm o oi o as vo vo vo VO 5-6 5-7 .v M _o sc — 0 <3 c < o £ ■'st 04 co 04 os jc . o C < o £ -g O vo >0 >0 io vo T+ 03 ^ ^vo g vo vo <3 Cj to z MOM ^t 04 OS 04 z 04 VO — 04 as 04 > — i co (N m > •O t- c co : > 03 • 03 U.hU u 60 ^.£ 03 in > : > c3 • a3 U.bU 8 60 x-£ c2 g** g co So^ C u Q.._. (D 8 5 is 6 1 p ^ A§ 2 So '5 -n W 3 oda o.--. (U ■—■ «-■ a) a o3 *2 | o -a 1 h 03 C o 5 (U 3 oda PQ^O 2 QQ^O s 89 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. JD aJ to O in C/3 r* s z > <73 to c CD >-) O' O z JD D, £ 03 CO I Tf in > co co £ n in «n >n Tt- Tj- Tf ui o '— f" CL) r- r- C/3 £ «n •n o £ co (N LUUI S co in £ b O b o 43 £ ovmo r- os T3 . $3 00 Os OS £ ‘O vo VO in >n ■glx g tT Tt- tT Sh O »-< OV oo »n »n Tj- Tf 1 10 1 0 r- i — 1 Z hco-O Tf in > VO Tt- £ co Tt - 1 Tf CO E o o £ ** b o £ •n't O Tf LUUJ X [AI qi Tf co (N £ t- t-* r- r- r- r- 60 c CL) oo r- «n -1 ov r-' £ r- r- r- b £ r- r- £ r- vo in o e OO 00 vo r- vo VO vo Tj- > CO > > a cs o3 cS UbUS -c .£ U.bUS ^ a £ C -1 c /3 60 ^ P-'T •“ . — <— CD D 03 +3 o^a CL) > 33 43 ID (U aS *3 QQ^O QQ >0 r» co 6 Devil’s Lair, all together/ Deepdene Cave, all together 1-47 38 Length, M 4 Deepdene Cave, all together/Devil’s Lair, all together 0-66 21 91 Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. Table 9 Comparison of tooth dimensions ( mm ) in specimens of macropodid species often found with Petrogale with mean dimensions of south western fossil Petrogale P 3 dP 4 P 4 M l M 2 M 3 M 4 Catalogue No. 1 x p.w. 1 x p.w. 1 x p.w. 1 x a.w. 1 x a.w. 1 x a.w. 1 x a.w. x p.w. Petrogale — (mean) 5-2x 3-3 5-5 x 4-4 6-8x40 5-8 x 5-0 6-5 x 5-6 71x6-2 7-5 x 6-3 x 5-4 Macropus irma — 77.6.496 5 -7 x 3 -8 64x5-2 6-5 x 5-7 7 -2 x 6-3 7-7 x 6- 1 M6788 6-3 x 3 • 8 6-7 x 5-8 7-3 x 6-4 81x6-6 8-5x60x5-5 Macropus eugenii — 65.10.1 15a — x 2-6 4-3 x 3 -7 4-7 x 2-5 50x4-3 5 -4 x 4-5 (P 4 ex- tracted) 69.3.781 4-3 x 2-8 5-5 x 4-9 5 • 7 x 5 • 1 6-2 x 5 -4 6-8 x 5-5 x 4-6 69.3.776 4-8x30 4-8 x — 5-2x 4-3 5-7 x 5-2 6-7 x 5-4 Setonix brachyurus — 77.6.323 7 -2 x 3 -5 4-2 x 4-2 4-7 x 4-5 51x4-8 5 1x4-4x3-5 77.3.449 4-9 x 3 -5 41x4-0 4-4x41 P 3 dP 4 P, Mj m 2 m 3 m 4 Catalogue No. 1 x w. 1 x p.w. 1 x p.w. 1 x a.w. 1 x a.w. 1 x a.w. 1 x a.w. x p.w. Petrogale — • (mean) 4- 7 x 2-3 4-9 x 3-3 6-2 x 2-4 5-6x3-5 6-3 x 4-2 70x4-6 7-3 x 4-9 x 4-4 Macropus irma — 70.12.458 49x3-2 61x4-4 6-7 x 4-6 7*2x 5-1 7-8 x 5-8 M6788 4-5 x 2-6 6-3 x 4-7 6-9 x — 7-6 x 5-5 7-4x 5-5 x 5-0 66.2.114 4-7 x 2-5 6-0x — 7-4x — 7-7 x 5-5 6-9x5-2x4-8 Macropus eugenii — 5-6x41x3-9 73.10.1354 3 -8 x 1-9 4- 7 x 3 -2 4-7 x 3 -5 5 • 4 x 4 • 1 65.10.115 3-7x 1-8 4-5 x 2-8 3 -5 x 1-5 4-7x31 5-0x 3-5 (P 4 ex- tracted) Setonix brachyurus — 5 -4 x 4-2 x 4- 1 10754 73.7.571 4-5 x 2-4* 3-6x 2-8 61 x 3 -0 4-6 x — 4-8 x 3-8 51x4-2 70.12.80 3 -9 x 2-7 4-3x31 * Maximum width of P 3 is anterior. 92 Table 10 Measurements {mm) of anterior cheek teeth o/Petrogale from various layers in Devil's Lair excavations of 1973-76. Left side only. Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. vo t — VO 00 t — ’sD co r- £ co co co co co co co co 5 d X X X X X X X X X Os vo vo oo vo O vo ON I/O vo vo vo vo vo vo vo VO VO vo co < Ui So o3 hJ O o -H -H -H o vo o o V O ON VO : ok a Table 10 — continued. Journal of the Royal Society of Western Australia, Vol. 61, Part 3, 1979. 0- u, > 03 (N (N I m m XXX •n r-~ •ccn so m (N > m m m m m m m PH d X XX X X X X X oo OS so n «n •n * r-* oo os O oo > m m m •'f rf rf rf rf rf rf rf rf rf rf Ph d X X XX X X X XXX X r- so r- in rf i 1 r >n in m m m m Ph d XXX X X XXX X X X X X X X •n X £ d T> w -a Ntn-dd--; X X X X n oo Tt «n »n »n in »n in T-H «n «n «n Tj- «n •n m -h oo O (N n-Hinoo £ m m m m m £ (N fN G fN fN fN — 2 L a XXX X X cC a X X o X X X X X m fN m O Tj- X o in £ Tt — i m r- VO vo VO t-- vo vb •n vo vo vo «n ON ov Tf VO fN OV • fN fN • £ • • • • m £ • • fN tT "d- Tt m o m m o CU Q- X X X X X CU a XXX "3 X Tf VO Ov 00 fN 73 X VO VO t*** in in^t Tfin T d - in m ov in © h rn fN £ m m fN m m £ fN fN fN X X X X X X XXX X fN Tt ov in O fN vo m «n «n »n «n Tf Tf tT : : : o ; o ' L- CO -D T3 O . . ; o ; : : : : O : : : • • o • • ■ • • • •••• 6 Ih c . . c . . . : cd 6 G G cd : & : : Z £ Z c * c a cd a> cd cd ctf g ^ee : *■* •£ £ :VJ.S o £ > a.2 g g CJ cd u. •— U Z *> at, OC o a U > o. c 5 x: g =o r\ ^ u -P 1 1 Ov \ > HH C o CO H cd « C S.S-g-S 8 3 & \ i fN 1 vo 1 1 1 vo in ^ * ov 1 0) a- I 1 ov m 1 1 Tt ov ov * ov I oo Tt • ov • OO OO fN 1 oo Tfr •1—1 • OO OO fN 1