The differences between the two regressions (Figs. 7) are due to the effects of seven regions. WP and TORR have only single species, while PILB, BARK, and BULL all have low richness (< thirteen spp.) (Fig. 6). These all are within the most arid portions of Australia, hence lack of water is the likely cause. LEB is also in the arid zone but has high richness (thirty). Two factors account for this. Springs, the groundwater source for which originates outside the arid zone, (hence aridity has no effect on water persistence), support eight endemic species. Secondly, LEB is much larger than most regions. When subdivided into individual drainages (which under present-day conditions are largely isolated from one another and areawise would be more similar to other regions), richness would be lowered to one through eleven for nine drainages and sixteen for one drainage when spring endemics are excluded (Unmack, 1995). Hence, if the geographic area was reduced, richness would be simil ar to other arid regions. The final odd value is for MDB. This region has a high latitudinal extent (ca. 12o), is around twice as large as PILB, the next smallest region (only WP and LEB are larger), and six times larger than all regions on a verage. While latitude per se does not explain richness, it is notable that mean annual rainfall is higher in the north, lowest in most of central Australia, and high again in southern-most latitudes (Fig. 4). Mean January maximum temperature shows a tr end of warm in northern-most areas, hot in middle areas, and coolest in the south (Fig. 5). Hence, the broad north-to-south climate trend is warm and wet, hot and dry, to cool and wet.
Comparisons among analyses
Overall results among clustering techniques are similar despite problems with the dataset in differences in the number of particular character states in OTUs, the small number of characters relative to OTUs, and the bias associated wit h each method. Cophenetic values were high for both clustering and ordination, hence both appeared to represent relationships equally well. Qualitative examination of Figs. 10 and 13 show five regions (CYP, BURD, FITZ, SEQ, and MDB) accounting for much of the variation in clustering results. As shown by ordination (Fig. 12), each of these regions is somewhat of a gradation between extremes rather than forming any discrete groups, or in the case of MDB of composite origin from faunally different regions (SWV, SEQ, and others). The inability of dendrogram clustering to represent regions of composite origin is a methodological limitation, as regions are forced into discrete clusters and hence any continuum of gradual change or multiple origin cannot be c learly represented.
Barriers to fish movement
Before discussing biogeographic divisions it is pertinent to review what barriers exist in Australia, their characteristics and timescales of influence, and provide examples. Four kinds of barriers are considered important: sea-water, drainage divides, climatic, and ecological.
Changes in sea-water barriers occur on both short- and long-term scales. Short-term changes in sea-level of perhaps up to 100-150 m occur in conjunction with short-term climatic fluctuations, probably on the order of every 100-150 Ky. Causes of these changes appear due to water storage in continental glaciers (Partridge et al., 1995). Long-term scales involve primarily three factors, change in the capacity of the oceans, change in continental elevation (Partridge et al ., 1995), and movement of lowland river channels such that their point of discharge into the ocean changes position (e.g., Nott, Idnurm & Young, 1991; Spry, Gibson & Eggleton, 1999). Another short-term means by which fish have been suggested to m ove between drainages is via riverine flood plumes (Williams, 1970; Chenoweth & Hughes, 1997; Jerry, 1997). These plumes may extend for sufficient distances to potentially connect adjacent drainages (Wolanski & Jones, 1981; Grimes & Kingsford 1996). However, the salinity and characteristics of these plumes can vary considerably and there are no data regarding fish occurrence within these plumes (Kingsford pers. comm.). Clearly this avenue of dispersal requires investigation.
In order to assess the importance of short-term sea-level fluctuations, drainage patterns were reconstructed to -500 m (Fig. 14). While -500 m is lower than any short-term sea-level known to have been attained, the results are more readily visualized; -100 and -200 m contour intervals are also shown. For some areas, lowered sea-level connects major drainage areas, e.g., between Australia and New Guinea, VIC and TAS, and Cambridge Gulf (WA and NT). Areas presently offshore of FITZ and possibly northe rn PILB would experience far higher connectivity than today. However, the remaining drainages are largely unaffected by short-term sea-level change.
Drainage divides are likely broken only over long time periods, primarily via drainage rearrangement through tectonism (see Introduction). Another possibility includes passage over low divides by fishes without rearrangement, i.e., th ey swim over. Two types of divides are identified, those between rivers with non-adjacent outlets (i.e., draining opposite directions), and those between rivers with adjacent outlets (i.e., draining the same direction with neighboring river mouths). The principal difference is their direction of flow relative to drainage direction. Those in adjacent drainages have divides parallel to direction of flow and those in non-adjacent drainages have divides perpendicular to direction of flow. Lateral channel migration is more likely to affect parallel divides than perpendicular ones. Parallel divides may allow movement of both upland and lowland species, whereas perpendicular divides only allow movement of upland species. Furthermore, adjacent river outlets have an additional means in their lowermost reaches by which fishes can be exchanged via lowered sea-level. If not adjacent, this is far less possible.
When comparing similarity across divides between non-adjacent inland regions (LEB and MDB) and coastal regions surrounding them, and also between adjacent coastal regions, it is clear the faunas of adjacent drainages are far more similar than in non-ad jacent ones (Fig. 15). Hence, given this higher faunal similarity I conclude connectivity occurred more frequently, or at least more recently, between adjacent drainages.
Climatic barriers occur along similar time frames as sea-water barriers. Regular, short-term fluctuations are on the order of 100-150 Ky, while long-term trends are also known (see Introduction). Climatic barriers differ from sea-wat er ones in being less sharply delineated and differing more in their effects between species, depending upon ecological tolerances (see below). The principal climatic barriers are minimum and maximum water temperature (which are determined by a combinati on of solar radiation, atmospheric temperature, and humidity) and rainfall. The first two determine physiological survival, the last permanence of water.
Due to its broad latitudinal and longitudinal area, Australia experiences considerable climatic differences between regions (Figs. 4 and 5). The most obvious climatic barrier is aridity, such as occurs in parts of southern, central, and western Austra lia where surface runoff is negligible. This has completely isolated SWWA, PILB, WP, TORR, LEB, and BULL from sea-level connections with surrounding regions since the last four are endorheic and the first two are isolated by WP. This limits potential fo r connectivity between regions and ones surrounding them to passage over divides. In northern, eastern, and southeastern Australia, climate is less extreme. Climatic change tends to be oriented north-south (Figs. 4 and 5). Climatic differences across n orthern Australia may have been bypassed during lowered sea-level as both adjacent and non-adjacent drainages became continuous at more northerly latitudes (i.e., Gulf of Carpentaria, Cambridge Gulf; Fig. 14). Hence, while climatic differences exist, the y likely had less effect on fish movement east to west. Although during glacial maxima, when sea-level was lowest, climate became warmer and drier in tropical areas (Williams, 1984), reducing potential for movement.
Considerable differences in climate exist along eastern drainages (Figs. 4 and 5). Northern regions (SECYP, NEQ) tend to have moderate rainfall and warm temperatures, with NEQ also having areas with high rainfall and cool temperatures due to higher el evation (Fig. 3). Mid-QLD regions (BURD, FITZ) are warmer and drier. This gradually ameliorates into progressively wetter and cooler conditions in the south.
Three lines of evidence suggest climatic effects have had the greatest impact on fish occurrence along the eastern coast, all the way from CYP south to SEV. There is a complete lack of distinctive faunal breaks, even within regions, as faunal differen ces between drainages are gradual and indistinct. If distinctive physical barriers existed, one would expect to see disjunction(s) as in other parts of the world where barriers have been identified (e.g., Obregón-Barboza, Contreras-Balderas & Lozano-Vilano, 1994). Also, disjunct populations of several species (detailed below) occur between regions or drainages with intervening areas being distinctively drier. Finally, fossil occurrences (Scleropages aff. leichardti and †Macq uaria antiquus) further south and north respectively of their descendant's present-day ranges, lend support to the climatic hypothesis. Areas where climate appears to have had little impact include southern VIC (SEV and SWV) and TAS (STAS and NTAS). Climatic differences between each pair of adjacent regions are negligible, yet several species occur in SEV that are lacking from SWV (see below) while NTAS and STAS have no species in common.
Ecological factors can be important when considering biogeographic patterns (Endler 1982). Ecological requirements (tolerances) are unknown for most Australian freshwater fishes, but tend to be broad for most due to uncertain habitat permanence under high within- and between-year (and longer term) variations in climate. Species with narrow tolerances are restricted to more permanent (i.e., more mesic) areas. Extinction probability is higher for a narrow habitat specialist unless it also has high migratory abilities; such abilities seem better developed in habitat generalists. Whatever the case, biology is inextricably interwoven with climatic, topographic, and other non-biological factors. The interplay among them will be sorted o ut only with accumulation of more ecological data than now exist. I leave that to the future.
Distributional relationships among regions (defined largely by the arbitrary hydrological scheme of Australian Water Resources Council (1976)) were compared in a search for broader patterns applicable to a landmass the size of Australi a. Major factors influencing fish distributions should be most evident at a larger scale, which I term "biogeographic provinces" as defined by Brown, Reichenbacher & Franson (1998); "… provinces are regional areas having a distinctive recent evolutio nary history and hence a more or less characteristic biota at the species and subspecies levels." For Australia, the evolutionary history may not be recent, and the "characteristic biota" may include differentiation to the generic or family level.
Designation of provinces often proves controversial, largely due to arguments about criteria (Horton, 1973), e.g., what percentage of endemism and number or percentage of different species is sufficient to separate one province from another, and how sp atially identical should species ranges be? I follow Keast's (1959) view, "… whilst it is legitimate to use the zoogeographic sub-region [=province] concept within the Australian continent it should be used only in broad context. … To think of it to an y degree in a static or absolute sense is quite misleading."
Opposing features of distinctiveness vs. similarity were quantified by examining species' occurrences among regions. Inland and surrounding coastal regions are shown from a northern coastal region, through inland Australia and then north to south alon g eastern coastal regions in Fig. 15, southern ones are shown west to east in Fig. 16, and northern coastal regions are presented sequentially from northwest, to northeast, to southeastern Australia in Fig. 17. Each number in the figures represents the d istribution of a species, with continuous lines indicating the presence of a species in adjacent regions. For each region, beginning from one direction, all known species are listed in sequence from those with the broadest range to those which are endemi c. Once a region has no more, new species are added sequentially in the next, those with the widest ranges first. In this manner, regions were combined into biogeographic provinces based on highest percentages of endemism supported by the greatest numbe r of species' ranges ending at a given boundary. Provinces resulting from hierarchical analyses (Figs. 10-13) were almost fully congruent with provinces hypothesized a priori through qualitative inspection of distributional data.
The following regions may be designated or combined into provinces (Fig. 18). SWWA and STAS (the Southwestern and Southern Tasmanian provinces respectively) each are distinctive since both have 100% endemism (Fig. 8) and as a result they do not group with other regions (Figs. 10 and 12). PILB and LEB each has a high percentage of endemics, 42 and 40% respectively (Fig. 8) and hence are designated Pilbara and Central Australian provinces. PILB does not group near to any other region (Figs. 10 and 13) , hence supporting its distinctiveness. LEB has close relationships to BULL and BARK (Figs. 10, 11, 13 and 15), and tentatively TORR; hence they are included with LEB in Central Australian Province. WKIM and EKIM contain seven and six endemic species of twenty-nine total for each, 24 and 21% respectively (Fig. 8). When combined, this increases to sixteen endemics (9.5% of Australia's freshwater fishes) out of thirty-nine species total (41% endemic) (Fig. 17). Their close relationship is also borne out in most results (Figs. 10, 12 and 13), justifying their combination as Kimberley Province. SWV and NTAS have endemism values of 22 and 29% (Fig. 8). NTAS has two endemic species, the remaining five are all in common with SWV and SEV. However, substant ial difference occurs between SWV and SEV in what is lacking in SWV. Six species occur in SEV but not in SWV, of the non-endemics in SWV all occur in SEV (Fig. 17). Hence there appears to be a unidirectional or superimposed recent barrier between these regions. Given this distinctiveness, and their combined endemism of 36%, and their consistent grouping together (Figs. 10 and 12) SWV and NTAS are combined into Bass Province.
MDB has 31% endemicity (Fig. 8) and complicated relationships with several surrounding regions (see below) as is demonstrated by its variable position between analyses relative to other regions (Figs. 10, 12 and 13). Among these, SAG has complete simi larity to MDB (Fig. 15), and Kulczynski's #2 Coefficient groups them together (Figs. 10 and 11), hence it is included in the Murray-Darling Province. Other surrounding regions are not included as the contribution of each is relatively small (see below). WP is categorized as the Paleo Province by default since it has only one species with no clear relationship to any one specific province or region.
Coastal areas from VOR east and south to SEV are difficult to categorize into provinces. A gradation of species ranges exists (Fig. 17) with no regional boundary being particularly distinctive and low endemism within-regions (Fig. 8). Clearly, the fa una of northern Australia is very different from that of the southeastern coast (Figs. 12, and 17), hence some degree of sub-division seems appropriate. The greatest difference in species' ranges between adjacent regions anywhere around the coastline occ urs at the boundary between NEQ and BURD where thirteen species (disregarding those endemic to NEQ) have their southern-most termini; many are congruent at the boundary. Hence, it seems reasonable to propose a Northern Province from VOR east to NEQ. Wit hin this province, endemism rises to thirty-eight species (50%) (Fig. 17), although twenty-five of these "endemics" and thirty-four species total also occur in southern New Guinea (recalculated from Allen (1991)), emphasizing its close faunal relationship with Australia. Clustering results support the distinctiveness of Northern Province (Figs. 10 and 11), and parsimony analysis also generally supports it, but with a slight incongruence involving SECYP and NEQ (Fig. 13). Three distinctive sub-provinces exist within the Northern Province. The first is referred to a Speciose Northern Sub-province, comprised of DALY, ARNH, EGC, ARCH, and CYP, with high species richness (Fig. 6 ) and few differences in faunal composition except four species present in DALY and ARNH not shared with the others (Fig. 17). The second, termed the Depauperate Northern Sub-province, consisting of VOR, WGC, NICH, and SGC, is notable for its lower species richness (Fig. 6). Finally, a Northeastern Sub-province consists of SECYP a nd NEQ, distinctive relative to the other two sub-provinces in lacking several species (Fig. 17). Further, NEQ has six endemics (Fig. 8).
Finally, an Eastern Province is proposed along the east coast from BURD to SEV. Despite there being no species in common at the extremities, no distinctive breaks are present between regions of this province. This north-south gradation is evident eve n at the drainage scale. Endemism within this province is 31% (Fig. 17). Ordination demonstrates the gradation between regions within the province most clearly (Fig. 12). Other results are mixed; clustering has an incongruence between BURD and FITZ and the other regions (Fig. 10) while parsimony analysis has three incongruences (Fig. 13).
Patterns of relatedness and their causes among provinces
Southwestern Province (SWWA): This province has no species in common with any other, suggesting long-term isolation. One family, Lepidogalaxiidae, and two percichthyid genera, Bostockia and Nannatherina, are endemic. At the generic level its highest similarity is with eastern regions (Table 1). SWV shares three of seven genera (Galaxias, Galaxiella, and Nannoperca), while MDB is the nearest region also containing Tandanus. There are no fami lies in common with northern regions. Faunal relationships clearly lie with southeastern Australia, although only distantly.
No fishes are recorded north of Southwestern Province until the Greenough drainage in PILB, nor east until TORR/SAG (with one possible exception, see Paleo Province). Mean rainfall decreases northward, while mean January maximum temperature rapidly in creases (Figs. 4 and 5), making conditions too dry for survival. Within this province, fishes are largely restricted to areas with mean annual rainfall >700 mm.
Chilcott & Humphries (1996) suggested Galaxiella may have migrated east-west across southern Australia as recently as Late Pleistocene. Given the aridity and sedimentary record across the Nullabor Plain this is unlikely. Conditions would h ave been favorable during Late Eocene when rivers supplied abundant sediments, major dune systems formed along the coast, and climate was temperate (Benbow, 1990). Several transgressions of the sea in this region from Eocene to Mid-Miocene also potential ly enhanced dispersal opportunities along coastlines (see Introduction). The Nullabor Plain formed 14-16 Ma with final regression of the sea from Eucla Basin about the time paleodrainage development began across southern Australia (Van de Graaff et al ., 1977; Benbow, 1990). It lacks any signs of integrated surface drainage due to the fractured nature of its limestone, which allows most surface water to seep underground rather than run off (Jennings, 1967; Benbow, 1990). Hence, Miocene would appe ar the minimum age when fishes could last migrate east-west across southern Australia.
Pilbara Province (PILB): Five of twelve species are endemic (Figs. 6 and 8), however their relationships to other species are poorly understood. The remainder comprise the most widespread species in Australia, found eastward around the coast to at least BURD, including LEB (Figs. 15 and 17).
Pilbara Province is one of the hottest areas of Australia. Mean January maximum temperatures are ca. 42oC in parts (Fig. 5). Mean annual rainfall is <400 mm (Fig. 4). Perennial surface water is scarce, and mostly persists in gorges. P ilbara Province is separated from Kimberley Province by the Great Sandy Desert (Paleo Province), an area with no surface runoff. No data exist to provide an indication of when this paleodrainage developed and hence isolated the two provinces from each ot her. The only Tertiary sedimentary records, consisting of minor alluvial and lake deposits (Taylor, 1994), are undated.
Kimberley Province (WKIM and EKIM): Endemism is high in this province. In the family Eleotridae, six of nine species are endemic (one endemic genus); Terapontidae, five of nine (one endemic genus); Atherinidae, two of three; Melanotaeniidae, tw o of five; and Toxotidae, one of two (Table 1). Several endemics have ranges limited to one or two individual rivers. Only sixteen species are widespread to at least east-coast drainages (Fig. 17). Three have more limited distributions east to DALY, th e remaining four occur across northern Australia between ARNH and CYP. Two species in EKIM have disjunct isolated populations, Melanotaenia nigrans (Fig. 17) and M. exquisita [which is absent between the King George (EKIM) and Daly (DALY) r ivers except one record in Pentecost drainage (VOR) (B. Hansen pers. comm.)]. It is likely other relictual populations of similarly distributed fishes may be found in EKIM, e.g., Pseudomugil gertrudae, P. tenellus, Denariusa bandata, and others. Three species are present in Fitzroy drainage (WKIM), occurring eastward through the Northern Province (Fig. 17), but are absent from the remainder of Kimberley Province. They include Arius midgleyi, Anodontiglanis dahli, and Craterocephalus stramineus. Glossogobius sp. C occurs as an isolated population in Prince Regent drainage, the nearest records to it being in the Pentecost drainage.
Several explanations exist for the high percentage of endemics. The province contains rugged topography with more gorges than any other part of Australia. Gorges provide refuges for fish during dry periods as they force hyporeic water to the surface. Bathymetic data provides conflicting evidence. A shallow submerged ridge exists to the north of the boundary between WKIM and EKIM, which could have been an important drainage divide during low sea-levels (Fig. 14), limiting opportunities for dispersal and thus increasing isolation. However, it does not explain, and in fact contradicts the close faunal relationship between WKIM and EKIM and the lower faunal similarity between Kimberley and Northern provinces. Perhaps two patterns exist, an older rela tionship between EKIM and WKIM (e.g., Craterocephalus lentigenosus, Syncomistes trigonicus, and Mogurnda sp. 1) and a more recent one with Northern Province (e.g., Neosilurus ater, Melanotaenia exquisita, Melanotaenia nigrans, Ambassis macleayi, and Syncomistes butleri) (Table 1, Fig. 17). Further phylogenetic work may help clarify this hypothesis.
Northern Province (VOR, DALY, ARNH, WGC, NICH, SGC, EGC, ARCH, CYP, SECYP, and NEQ): The western boundary of this province is marked by the species it lacks relative to regions to the west as outlined under Kimberley Province. The southern boun dary has the largest absolute decline in number of species present, from thirty-nine to twenty (Fig. 6) and is the southern limit for thirteen species, the highest change anywhere along the coast of Australia (Fig. 17). Regions within the Northern Provin ce have fourteen endemic species (Fig. 8), four of which are also shared with New Guinea (Allen, 1991). When the whole province is considered, there are twenty-four additional endemics, raising the total to thirty-eight (twenty-five of which are shared w ith New Guinea) out of seventy-five species (50% endemic) (Fig. 17). Within this province, three sub-provinces may be recognized: Speciose Northern, Depauperate Northern, and Northeastern Sub-province.
Speciose Northern Sub-Province (DALY, ARNH, EGC, ARCH, and CYP): This sub-province is the richest of the province (Fig. 6), with its own distinctive fauna (Fig. 17). Several taxa occur on each side of the Gulf of Carpentaria in the Speciose Nor thern Sub-province, but not in the Depauperate Northern Sub-province, e.g., Anodontiglanis dahli, Porochilus obbesi, Iriatherina werneri, Melanotaenia nigrans, M. trifasciata, Pseudomugil gertrudae, P. tenellus , Denariusa bandata, and Oxyeleotris nullipora. An exception is Hephaestus carbo which occurs as an isolated population in NICH (Depauperate Northern Sub-province ) as well as in the Speciose Northern Sub-province on each side of the Gulf.
The causes of this trend in richness appear climatic. There is a substantial rise in annual mean rainfall from 600 mm to >1000 mm from southern portions of the Depauperate Northern Sub-province to the Speciose Northern Sub-province (Fig. 4). There also is a corresponding decrease in mean January maximum temperature from the upper 30's to the low 30's (Fig. 5). Mean July minimum temperature also increases from 12-15oC to >15oC.
Depauperate Northern Sub-province (VOR, WGC, NICH, SGC, and EGC): There is a decline in richness from the Speciose Northern to the Depauperate Northern Sub-province (Fig. 6); i.e., thirty-nine (ARNH) to thirty (VOR) and twenty-seven (WGC); and t hirty-six (EGC) to twenty-nine (SGC). This sub-province is characterized by lower annual rainfall, higher summer maximum temperatures, and lower winter minimum temperatures, as noted above. With the partial exception of NICH, which is fed by large peren nial springs, most regions lack several species. Furthermore, several species are rare and patchy in occurrence, but are otherwise widespread and common in other sub-provinces. Mogurnda mogurnda occurs in VOR and NICH, but is only known from a si ngle record in SGC, and from two drainages in WGC. Craterocephalus stercusmuscarum is only known from a few small collections. Hypseleotris compressa is recorded from only VOR and three other drainages (Limmen Bight (WGC), Leichardt, and N orman (SGC)). There are several drainages from which Glossamia aprion is recorded, however, it is far less widespread or common than in northern ones. Few records of Ambassis agrammus and A. mulleri exist, while Scleropages jardi nii is only recorded from Roper drainage (WGC) and NICH and Hephaestus carbo is only known from NICH.
Within the Depauperate Northern Sub-province, Pingalla gilberti is the only endemic, while Scortum barcoo and Porochilus argenteus are not found in the other sub-provinces (Table 1). The latter two are typical of the Central Austr alian Province and I speculate they are excluded from northern drainages by competition with tropical-adapted fishes.
Northeastern Sub-province (SECYP and NEQ): A significant drop in richness occurs from CYP to SECYP (forty-five to thirty-five) (Fig. 6) and several species have their eastern range limit here (Fig. 17). These changes may be due to several facto rs, including lower mean winter minimum temperature, and a lack of major rivers due to closeness of the Eastern Highlands to the coast (with exception of Normanby drainage).
NEQ has both high richness (thirty-nine) (Fig. 6) and number of endemics (six) (Fig. 8). This is an unusually high value for richness compared to surrounding drainages and for endemism compared to drainages east of the "endemic line" (Fig. 8). Severa l interrelated factors likely account for this. The Eastern Highlands are considerably higher (500 - 1622 m) in NEQ than any regions to the north and southward to FITZ (Fig. 3). Although the area of high elevation stretches beyond NEQ, only ranges near the coast experience high mean annual rainfall (>2400 mm); rainfall decreases to <1200 mm a short distance inland (Fig. 4). Summer maximum temperatures also are lower due to higher elevation (Fig. 5). Hence, it provides a refuge with considerably higher rainfall and cooler summer maximum temperature to species likely more widespread when climate permitted. Further evidence for a refuge can be inferred from other fish distributions. Several species have their southern-most limits here, although s ome are not continuously present northward. For example, Oxyeleotris aruensis and O. fimbriata occur as isolated southern populations, while Melanotaenia maccullochi, Pseudomugil gertrudae, Denariusa bandata, and Oxy eleotris nullipora occur in patches from CYP south into NEQ.
Northern Province and Southern New Guinea: A strong relationship exists between Fly River, New Guinea (as well as most of southern New Guinea) and the Northern Province of Australia, with thirty-four out of seventy-five freshwater fishes (45%) i n common (Roberts, 1978; Allen, 1991). Furthermore, four species, Nematalosa erebi, Amniataba percoides, Craterocephalus stercusmuscarum, and Hephaestus carbo (and likely others, yet uninvestigated) have apparent sister specie s in Fly River (increasing commonality to 51%). Presently, Fly River drains to the Southern Pacific Ocean. However, it is hypothesized to have been diverted from a southern route directly into Arafura Sea (Blake & Ollier, 1969) between 35 and 40 Ka by upwarping (Torgersen et al., 1988). The Arafura Sea between Southern New Guinea and the Northern Province is shallow, hence regularly exposed during lowered sea-level, which would potentially connect most drainages in Torres Strait west of Cape York Peninsula (Fig. 14).
When did the fish faunas of Australia and New Guinea last meet? Allen & Hoese (1980) suggested severing of the connection was 6.5-8 Ka during the last sea-level rise. While likely correct for hydrological connectivity, it seems unlikely for most freshwater species as a migration route. During Late Pleistocene, a major lake known as Lake Carpentaria existed east of Groote Eylandt during lowered sea-levels (Fig. 14) (Torgersen et al., 1983; Jones & Torgersen, 1988). It is not known whe n the lake first formed, but, for much of its examined history (up to 40 Ka), it was brackish or fresh-to-brackish (Torgersen et al., 1988), which may have prevented migration. Furthermore, during lowered sea-level it is thought this area was more arid than at present (Webster & Streten, 1972), hence possibly decreasing available aquatic habitats. Major faunal exchanges are more likely to have last occurred during low sea-levels during Late Miocene (Haq et al., 1987) when global climat e was warmer and wetter than today (Partridge et al., 1995).
An interesting aside is the absence of five of the six most widespread Australian fishes in New Guinea. Leiopotherapon unicolor and Neosilurus hyrtlii have no known near relatives in New Guinea, Nematalosa erebi and Amniataba p ercoides each have likely sister species (largely restricted to Fly River), while Melanotaenia splendida has a widespread allopatric subspecies in New Guinea (Allen, 1991). Given their widespread occurrence and broad environmental tolerances ( Merrick & Schmida, 1984), these species are most likely of all to have been able to migrate during the last low sea-level, but they are absent. One could speculate many possible reasons for this, i.e., competitive exclusion or incorrect taxonomy, alt hough insufficient evidence exists to warrant further discussion. There are also species with extensive southern New Guinea distributions that do not occur in Australia. Examples include Arius carinatus Weber, A. latirostris Macleay, A. macrorhynchus (Weber), the catfish genera Cochlefelis, Doiichthys, and Nedystoma, Zenarchopterus novaegunineae (Weber), Melanotaenia goldiei (Macleay), and Glossamia sandei (Weber). Furthermore, there are se veral species with more limited ranges in central southern New Guinea in the middle and upper Fly River that do not occur in Australia (Allen, 1991). Clearly, this region has a long complex history that is poorly understood.
Eastern Province (BURD, FITZ, SEQ, NEN, SEN, and SEV): The Eastern Province is distinctive for its lack of faunal breaks (Fig. 17), its boundary with the Northern Province being due to a sharp decline in richness (Fig. 6) and disappearance of th irteen species from NEQ to BURD (Fig. 17). No particularly distinct faunal breaks occur even at the drainage level until Wilson's Promontory at the southern boundary of the province (see below). Only five endemics occur at the regional scale (Fig. 8). Even when the entire province is considered, endemism only rises to fifteen out of forty-eight species (31% endemic) (Fig. 17).
The differences in richness and the pattern of species occurrence between regions within this province are likely due to climatic effects (Figs. 4 and 5). BURD and FITZ are both the driest and hottest areas on the east coast. From SEQ south to SEV, r ainfall steadily increases; however, mean maximum January temperature also decreases. The continental shelf is particularly narrow in southern regions, being broadest offshore of FITZ (Fig. 14).
Biogeography of Burdekin River (BURD) was recently discussed by Pusey et al. (1998), who attributed low richness to several factors including Burdekin Falls, substantial volcanic activity, past climatic stress, and low habitat diversity. The ar ea has experienced considerable volcanic activity, as have several regions in eastern Australia (Johnson, 1989). Also, times between eruptions, while not specifically documented, are likely sufficient to allow recolonization and minimize long-term impact s. Past climatic stress is difficult to infer, however, present-day stress is higher than for any other east-coast drainage (Figs. 4 and 5) and would likely have been more so in the past when climate was drier during glacial maxima (Williams, 1984). Mos t of the drainage receives <1000 mm of rainfall a short distance inland of the coast while southern parts receive <600 mm. Summer maximum temperatures inland of the coast are higher, and winter minimum temperatures are lower. Low habitat diversity was demonstrated by Pusey et al. (1998). I consider harsh climatic factors, combined with low habitat diversity as the most parsimonious explanations for low richness, although clearly the falls have excluded some species.
If Burdekin Falls is a barrier, other species should also be present today in surrounding drainages and below the falls, yet lacking from above them, whereas if climate were the cause, species would also not occur in surrounding drainages with similar climate. Pusey et al. (1998) demonstrates the falls are a barrier to species known to require estuarine or marine habitats for reproduction, and for three freshwater species; Pseudomugil signifer, Glossamia aprion, and Oxyeleotris lineolatus. However, it is unclear whether O. lineolatus is native below the falls and for the purposes here I assume it is introduced due to the lack of records from adjacent drainages (see below). Tandanus ta ndanus and O. lineolatus are known to be introduced above the falls and are either not present (Pusey, pers. comm.; Pusey et al., 1998), or considered introduced (Hogan, pers. comm.) below them respectively. Given both can survive prese nt conditions in the drainage as demonstrated by their introduction and apparent naturalization, the falls hypothesis would appear to be supported. However, they are absent (except one record of T. tandanus from Pioneer drainage (FITZ) which may a lso be introduced) from north of Fitzroy drainage (FITZ) until one encounters Murray and Herbert drainages respectively (NEQ), areas beyond the influence of Burdekin Falls. Hence, the climatic hypothesis is also supported since T. tandanus and O. lineolatus are also absent from surrounding drainages. Additional arguments presented by Pusey et al. (1998) in support of the Burdekin Falls barrier hypothesis include absence of Gobiidae and the lower richness of Eleotridae and Ambassidae relative to NEQ. Again, all these species lacking from Burdekin drainage are absent from Murray or Herbert drainages southward. Hence, evidence for both climatic and waterfall barrier hypotheses can be demonstrated.
J. Stephenson (pers. comm.) (also quoted in Pusey et al., 1998) suggested part of the upper Burdekin River formerly flowed into Gilbert drainage (SGC) prior to uplift of McBride Plateau ca. 8-10 Ma, prior to widespread volcanism. This diversion is largely inferred; direct physical evidence is lacking. Pusey et al. (1998) suggested this as a possible colonization route for Neosilurus mollespiculum (as sp. C), Porochilus rendahli, and Scortum as they are all "common" in Gulf of Carpentaria drainages (CYP west to ARNH) and uncommon in northeastern drainages. However, the nearest populations appearing closely related to N. mollespiculum are in BULL (the species remains unidentified there and may be distinct) and DALY, west to WKIM (N. pseudospinosus) (Table 1). It is unrecorded from Gulf of Carpentaria regions. Porochilus rendahli is widespread, occurring, albeit patchily and in low abundance, from Brisbane drainage (SEQ) northwards up the entire east coast and across northern Australia (Table 1). Scortum also occurs south in Fitzroy drainage (FITZ) and west in the Central Australian Province (Table 1). The phylogenetic position of Scortum parviceps, when defined, may help clarify relationships and hence history. Hence, definitive data supporting a western colonization route are lacking. Additional species listed as possible invaders include Amniataba percoides, Hephaestus fuliginosus, and Leiopotherapon unicolo r, although all are widespread (Table 1) and could have migrated around the coast. Furthermore, several species present in southern Gulf of Carpentaria regions are absent from east coast regions (Fig. 17, Table 1). If colonization did occur via this route, its "signature" may be been overwritten or confounded by events since; present evidence is equivocal.
FITZ is notable for its isolated southern population of Oxyeleotris lineolatus, coastal occurrences of Macquaria ambigua and Scortum hillii (Table 1), and disjunct northern populations of Rhadinocentrus ornatus, Pseudomug il mellis, and Gobiomorphus australis. A similar species tentatively identified as S. hillii is also found in EGC and possibly LEB (Vari, 1978). Other possible relationships could be either to S. parviceps in BURD or S. barco o in Central the Australian Province and Depauperate Northern Sub-province. Electrophoretic evidence suggests coastal M. ambigua are most closely related to those from MDB rather than LEB (Musyl & Keenan, 1992).
SEQ has a mix of northern and southern species, hence giving it the highest richness (Figs. 6 and 17). Eight species have their southern coastal limit at SEQ, a further five continue south via MDB (Fig. 15). Eight also have their northern-most occurr ence in this region (Fig. 15).
SEN is distinctive due to the occurrence of Macquaria australasica in Shoalhaven and Hawkesbury drainages; it is otherwise only known from MDB. Dufty (1986) suggested Shoalhaven and Hawkesbury populations may be separate species from each other as well as from MDB based on electrophoretic and morphological data.
Clearly, fish movement between adjacent coastal drainages is neither easy nor frequent along much of the Eastern Province. While sea-level and climatic changes are relatively frequent, few drainages actually connect during low sea-levels (Fig. 14). H ence, to move, fish must rely on long-term processes such as drainage rearrangement. Several species had historically broader ranges along the coast, now fragmented likely by climatic changes. Most species, in both their northern and/or southern range e xtremities, tend to have their termini independent of co-existing species, suggesting a differential "filter" again likely caused by climate. Evidence supporting a significant long-term (perhaps greater than a few million years) lack of mixing between ad jacent (or relatively close) drainages comes from studies on Tandanus tandanus (Musyl & Keenan, 1996; Jerry & Woodland, 1997), Melanotaenia duboulayi (Crowley, Ivantsoff & Allen, 1986), Maccullochella (Rowland, 1993) and Macquaria australasica (Dufty, 1986). Other examples demonstrate a higher level of gene flow; Pseudomugil signifer exhibits clinal variation along its range (Hadfield, Ivantsoff & Johnston, 1979) and Macquaria novemaculeata follow s an isolation by distance model with limited gene flow between adjacent populations (Jerry, 1997; Chenoweth & Hughes, 1997; Jerry & Baverstock, 1998; Jerry & Cairns, 1998). Both species spawn and/or occur in upper estuarine areas, and may ha ve a higher proclivity for dispersal between river mouths via riverine flood plumes due to higher salinity tolerance. It is likely Retropinna semoni, Macquaria colonorum, Gobiomorphus australis, G. coxii, Hypseleotris compr essa, Philypnodon grandiceps, and Philypnodon sp. will all show a similar pattern of minor variation between drainages as they all may also be found in upper estuarine areas. It is also notable that this group of species is dominant and widespread in the southern half of this province where the continental shelf is narrowest. Hence, for some species, the limited evidence suggests distribution patterns along the Eastern Province have been achieved over a vast period of time, while other s may be continuing to occasionally exchange individuals today.
Bass Province (SWV and NTAS): Bass Province is distinctive in what it lacks relative to SEV (Eastern Province). Both SWV and NTAS share five species with SEV, however six species present in SEV are absent here (Fig. 17). Each region has two en demic species, Nannoperca obscura and N. variegata (SWV) and Galaxias tanycephalus and Paragalaxias mesotes (TAS) (Table 1). Of the latter, the first is a land-locked species derived from the diadromous G. truttaceus , while P. mesotes is related to other Paragalaxias spp. in STAS. One species has its western limit in SWV, five have their southern limit in NTAS (Fig. 16). While the two regions are presently isolated by Bass Strait, it is shallow an d was fully exposed during low sea-levels (Fig. 14). Of note is the former drainage patterns during low sea-level in the vicinity of Wilson's Promontory (Fig. 14). The faunal disjunction between Bass and Eastern provinces occurs here, but the former dra inage divide appears to the east of it. The bathymetric reconstruction may be misleading, as that portion of the sea-floor is relatively flat making it difficult to predict drainage direction. Shifts in drainage direction would only require minor change s in topography. Of the four species whose ranges end at Wilson's Promontory, two (Macquaria novemaculeata and Philypnodon sp.) are common, and it is most likely they would survive in Bass Province. The other two (Gobiomorphus australis and G. coxii) are relatively rare in this portion of their range, hence may be limited by some ecological factor(s). Despite this incongruence, present day data demonstrate the faunal divide is near Wilson's Promontory.
Southern Tasmanian Province (STAS): This province has no shared species, all eight are endemic (Fig. 8). Three galaxiids, Galaxias auratus, G. fontanus, and G. johnstoni appear derived from the diadromous species G. brev ipinnis and G. truttaceus (Merrick & Schmida, 1984). Three Paragalaxias spp. occur in Central Plateau lakes. With the exception of Paragalaxias mesotes (NTAS), which almost definitely crossed the drainage divide, and possibl y G. parvus and G. pedderensis (STAS), no other fishes have dispersed to or from the Southern Tasmanian Province. The remainder have all likely evolved in situ. All have quite restricted ranges, often to one or a few lakes and/or st reams (Allen, 1989). This lack of dispersal is likely a combination of being at the extreme southern end of Australia, hence experiencing the coldest extremes during climatic fluctuations, and effective ocean barriers since the continental shelf is quite narrow.
Murray-Darling Province (MDB and SAG): SAG is combined with MDB to form this province as it shares all four species with it (Fig. 15). During low sea-level, all drainages south of Wakefield (SAG) exit southward via the Gulf of St. Vincent (SA) (northern drainages flowing into Spencer Gulf lack freshwater fishes). Present subsurface topography in the Gulf of St. Vincent would divert drainage to between Kangaroo Island and Fleurieu Peninsula, eventually joining the lower Murray (Fig. 14), hence explaining similarities of their fauna. This gulf is relatively shallow, however, minor changes in topography could easily divert drainage to either side of Kangaroo Island and hence away from Murray drainage.
Murray-Darling Province has complex relationships to surrounding regions. It has high similarity to SEQ (thirteen of twenty-eight species), SWV (six of nine species), and LEB (nine of thirty species) (Fig. 15). It is worth noting LEB and SWV have one species in common (it occurs in all southeastern regions), LEB and SEQ have nine (six of which are widespread northern species), and SWV and SEQ have two in common (Fig. 15). Additionally, Murray-Darling Province has nine endemic species (Fig. 8). Over all, it appears to have experienced mixing of faunas from different surrounding regions with distinctive faunas relative to each other, while also maintaining a high degree of endemism. With the possible exception of SWV (which is adjacent), all faunal c onnections must have occurred across drainage divide barriers.
It is not possible to assume fish distribution patterns of today are directly reflective of historical ones in as far as species ranges in Eastern Province. Some species may have occurred further south in Eastern Province than at present. The followi ng discussion should thus be regarded as tentative; several alternatives may exist.
A peculiar pattern in southeastern Australia is the occurrence of several species southward along the east coast of Australia to around Brisbane drainage (with some occurring southwards until Manning and Hunter drainages), then continuing further south via Murray-Darling Province although absent from other intervening southeastern coastal regions. A number of recent ichthyological papers (Crowley, 1990; Musyl & Keenan, 1992, 1996; Rowland, 1993; Waters et al., 1994; Pusey & Kennard, 199 6; Pusey et al., 1998; and others) have considered river captures across the Eastern Highlands common, based upon geological evidence, and hence used that explanation for this pattern. However, the dominant geomorphic paradigm has shifted to where river captures are thought rare, or non existent there (see Introduction). The Eastern Highlands (ca. 90 Ma) predates most of the fishes, hence they must have crossed it. The alternate possibility of once being widespread throughout southeastern Austra lian and dispersing via coastal drainages to subsequently become extinct, seems unlikely, especially since there are no gaps in species' ranges along the southeastern coast (Fig. 17). Other confounding difficulties in interpreting this history include th e possibility that fish may have crossed the Eastern Highlands at one site or several, on single or multiple occasions, and from alternating directions. Limited evidence suggests options for crossing are limited to only a few sites, however, the number a nd direction of exchange is going to be difficult to determine.
Several areas of lesser elevation occur in the Eastern Highlands along the boundaries between Eastern and Murray-Darling provinces. These include headwaters of the Clarence, Hunter, Hawkesbury (NSW), Snowy, Tambo, and Maribyrnong (VIC) drainages (Tayl or, 1911; Haworth & Ollier, 1992). Broad, low divides also are found between Fitzroy and Burnett drainages (QLD) and MDB and in eastern VIC between Hopkins and Glenelg drainages and MDB (Fig. 3). Upper parts of some of these have experienced volcani sm. Activity in Brisbane and Clarence drainages was from 22.6-27.2 Ma, Hunter from 31.8-42.7 Ma, Hawkesbury from 14.4-26.0 Ma, Snowy from 36.9-54.9 Ma, Tambo from 25.8-37.2 Ma, and Maribyrnong from 4.6-7.0 Ma (Johnson, 1989). Many of these dates are lik ely too old to have contributed to faunal exchange, but it could be modification due to volcanism subsequently made it easier for certain species to cross.
Faunal similarity clearly exists between drainages either side of several low places. Fitzroy drainage has Macquaria ambigua among eleven species in common with MDB, which is absent from other coastal drainages (Fig. 15). Clarence drainage has nine species in common with MDB (Fig. 15). Hunter drainage has six species in common, however, a record for Craterocephalus amniculus is more notable if it reflects natural occurrence in this drainage. I have considered it introduced due to its recency of collection (1976 and 1980, one and five specimens respectively) and lack of further records despite efforts to find it (Crowley & Ivantsoff, 1990). The Hawkesbury and Shoalhaven drainages are notable for their coastal populations of Mac quaria australasica; otherwise only found in MDB. They have five species total in common with MDB (Fig. 15). There is nothing unique on the coastal side of the low region in the Snowy, Tambo, or Maribyrnong drainages; however, five species are in co mmon. The same situation holds in other drainages in western VIC, where the divide is particularly low (Fig. 15).
Absence from coastal regions relative to MDB is demonstrated by Gadopsis bispinosus in upland streams in the Murray drainage (from the Goulburn River east to the Murrumbidgee River). Seemingly suitable habitat is unoccupied in coastal drainages . Gadopsis marmoratus also is absent from coastal NSW drainages, yet occurs over the drainage divide in MDB from Lachlan River north to Condamine River. Given their upland habitat both would be expected to have crossed the Eastern Highlands, one way or another. One species that may hold the key to identifying migration over the Eastern Highlands is Galaxias olidus. It occurs at all elevations in virtually all drainages along it on both sides, but is more frequently at higher elevations a nd in smaller headwaters than any other fish in southeastern Australia.
The relationships between LEB to the north and MDB also are obscure. Musyl & Keenan (1992) suggested MDB and LEB were connected during lower sea-levels via their southern exits to the sea when MDB "previously" flowed further westward into Spencer Gulf (SA) (Williams & Goode, 1978). However, the position of the southern outlet for LEB has never been identified, although one is hypothesized to have existed in Miocene due to the presence of freshwater dolphins (Tedford et al., 1977). Fur thermore, earlier suggestions of a more westerly outflow for MDB (Williams & Goode, 1978) have been rejected and its position considered unchanged since Paleocene (Stephenson & Brown, 1989). Others suggested LEB formerly had an outlet into MDB ne ar Broken Hill (NSW) (Tedford et al., 1977), although no supporting evidence was presented. Alternative connections may have existed via headwaters of Barcoo River and northwestern portions of MDB, or via Bulloo drainage (BULL).
Clearly the fauna of BULL is most closely related to LEB (Fig. 15), and several of its species have been tentatively classed as intermediate between those of LEB and MDB. Musyl & Keenan (1992) suggested Macquaria ambigua from BULL may demon strate hybrid influences between MDB and LEB stocks. Melanotaenia "splendida" appears related to populations of an undescribed Melanotaenia from the adjacent Paroo and Warrego rivers in MDB as well as M. s. tatei from LEB. Based on two specimens, G. Allen (pers. comm.) could not identify the Ambassis sp. from BULL, as they had intermediate counts between A. mulleri and A. agassizii, in LEB and MDB, respectively. Hence, although tentative, a connection via BULL between MDB and LEB appears likely. How or when this occurred is unknown.
Central Australian Province (BARK, LEB, BULL, and TORR): Four regions are amalgamated on the basis of their high faunal similarity (Figs. 10-15), and their most likely connections to other regions.
LEB has a complex relationship to surrounding regions. Due to its large size it borders many regions, which may account for its complex fauna. Its relationships with MDB have already been discussed. Nine species are shared between LEB and Fitzroy dr ainage of the Eastern Province, although one of these, Macquaria ambigua, is more closely related to MDB populations (see above), and another, Melanotaenia splendida tatei (of LEB) is considered more closely related to M. s. inornata from northern regions rather than M. s. splendida from the east (Allen & Cross, 1978). LEB and Burdekin drainage of the Eastern Province have eight species in common. In both cases, most species in common are widespread across northern and/or eastern Australia. Also, there is only a small segment of shared drainage divide between LEB and Fitzroy drainage. Overall, evidence thus is weak for eastern coastal relationships.
LEB has ten species in common with Flinders drainage (SGC), which, aside from seven widespread species also has Porochilus argenteus and Scortum barcoo in common. The latter two species only have a limited distribution in the Depauperate Northern Sub-province and Central Australian Province (Table 1). Connections between these provinces have been reported for the headwaters of Cooper drainage (LEB) and Flinders drainage via drainage rearrangement. Based on data and ages from volcanic o utflows upper Prairie Creek was rearranged into Flinders River between 5.5 and 3.3 Ma (Coventry, Stephenson & Webb, 1985). Also, the present headwaters of Diamantina drainage are thought to formerly have flowed into Flinders drainage based on geomorp hologic changes to the latter in the form of a lessened stream profile (Twidale, 1966). Unfortunately, no firm evidence exists as to when this occurred, but it was likely between 5 Ma to possibly Late Pleistocene (R. Twidale, pers. comm.). Hence, two co nnections, each in a different direction are identified. Overall, strong evidence, both biological and geological, exists for connections of LEB with northern Australia.
Presently, all drainage in BARK is internal, ending in intermittent lakes in the region's center. Relationships of fishes appear closest to LEB, based on presence of Melanotaenia splendida tatei in both. However, aside from this subspecies, al l other taxa are common to both LEB and WGC (Fig. 15). The eastern edge of BARK is barely separated from LEB in terms of elevation and heavy runoff might allow fishes to swim today between regions. Further, this low divide is downstream of the headwater s of streams in the area (Fig. 2), hence enhancing the potential for faunal exchange. Alternatively, headwaters of many drainages in WGC are on the northern edge of the Barkley Tableland, also barely separated in terms of elevational difference. It woul d only take slight tectonic shifts to change drainage directions. Overall evidence suggests the most recent connections were with LEB, hence its placement in that province.
Hydrologic associations of BULL with MDB have been suggested (Lake, 1971). While BULL contains a number of species in common with both MDB and LEB, it also has several species in common with LEB that are not in MDB; the reverse does not occur (Fig. 15 ) (although see discussion under MDB concerning hybrids). Furthermore, topographically BULL appears to have formerly flowed south of its present terminus, likely into Lake Frome (SA). My only evidence for this is slope of the valley, except for a slight rise over a low divide between Palgamurtie Creek (BULL) and Packsaddle Creek (LEB) between the southern end of Grey Range and northern end of Barrier Range (NSW) (Fig. 2). When such drainage last occurred is unknown, but is unlikely to have been since a ridity enveloped this area (perhaps since at least 1 Ma). Only one species in BULL is not found in surrounding drainages, an unidentified plotosid related to Neosilurus mollespiculum and N. pseudospinosus from BURD and northwestern regions, respectively. Presumably, this species was once more widespread and is a relict. Its true relationships are unlikely to be clarified unless further material becomes available.
Only one species is recorded from TORR (Fig. 6), and is tentatively identified as Craterocephalus eyresii, although this is likely incorrect. Craterocephalus eyresii was reviewed by Crowley & Ivantsoff (1990), resulting in separation into four species. Unfortunately, no material from TORR was examined. Its closest affinities are probably to LEB, given their close geographic proximity, and it is thus tentatively included in Central Australian Province.
Paleo Province (WP): Paleo Province contains former connections to surrounding drainages that were severed as aridity began in Mid Miocene (Van de Graaff et al., 1977). Six paleodrainage catchments exist relative to their former connecti ons to surrounding drainages: Paleo-Victoria, Paleo-Sturt, Paleo-Internal, Paleo-Oakover, Paleo-Southwestern, and Paleo-Southern (Fig. 18). Most of this province in NT formerly drained into Victoria drainage, northern and middle WA drained into Oakover d rainage and through the Great Sandy Desert, southwestern WA drained into Swan drainage, and those north of it in Southwestern Province, southern portions of WA and all of SA drained into the Great Australian Bight. Only a few records of Leiopotherapon unicolor are known from this vast region, which remains relatively unexplored in WA and SA. The only unusual record of this species (WAM) is from the southern draining Raeside Paleoriver (WA), which drains into the Great Australian Bight. It is yet to be confirmed if this record is valid, as only one collection exists. Former drainage connections within WP could have easily allowed northern fishes to migrate across Australia via an inland route, rather than around the coastal fringe. The former h eadwaters of Paleo-Victoria drainage are adjacent to those of Paleo-Oakover drainage, BARK, and LEB which clearly potentially enhances faunal exchange across a major area of Australia. However, when this may have last been possible is unknown since no de posits have been dated from the majority of this province.