CONCLUSIONS

The most striking pattern shown in the distribution of freshwater fishes across Australia is endemism. Most provinces in southern, central, and western parts of the continent have a large proportion of endemics, whereas Northern and E astern provinces, with exception of NEQ in the latter, have few. The pattern results from isolation due to aridity and drainage divides. Low endemism in Northern Province must be a result of high drainage connectivity during lowered sea level (Fig. 15), preventing isolation sufficient to promote local speciation. Low endemism in Eastern Province is more difficult to explain. Drainages appear to have been long isolated, even during lowered sea levels, yet faunal patterns suggest no distinct barriers an d only short-term isolation. Hence, climate seems the most likely cause of speciesí distributional limits.

Fossil evidence (Appendix I) demonstrates persistence of certain families and genera in freshwaters of Australia over at least the last 30-45 My. The record is strongly biased toward larger taxa, with most fossils of families containing species longer than 20 cm; smaller taxa are rarely found. Given the small size of many Australian species, their fluvial habitats where fossilization may be a rare event, and a general lack of small, whole fossils of any kind, this is not surprising. Crowley (1990) n onetheless suggested craterocephalids and possibly melanotaeniids have existed for a long time. Earlier hypotheses (Whitley, 1959; Allen, 1982; Merrick & Schmida, 1984; Williams & Allen, 1987; Allen, 1989) of the recent radiation of Australian gr oups in general seem unlikely.

Further, Plio-Pleistocene phenomena seem to have had little influence on expansions of ranges for most of the Australian Continent. Lowered sea-levels potentially connected SWV and NTAS, Gulf of Carpentaria regions and New Guinea, Cambridge Gulf regio ns, SAG and MDB, and drainages within FITZ (Fig. 15), however, the rest of Australiaís coastline remained essentially uninfluenced by sea-level changes. Examination of drainage patterns during lowered sea-levels clearly establishes many hypotheses for te sting. Some low sea-level divides occurred at boundaries between regions (i.e., BURD and FITZ, and FITZ and SEQ), while others do not match at all (i.e., bathymetry would predict EKIM to be similar to VOR and DALY rather to WKIM, and the boundary between SEV and SWV also does not match the presumptive drainage divide). Where widespread species cross faunal breaks, they should show discontinuities similar in degree of relatedness to populations in surrounding drainages if barriers are geomorphic rather t han climatic or ecologic.

Movements between drainages in regions not influenced by sea-level changes appear likely only over very long time-scales, especially given the geologic stability of Australia. Interacting with this process are both long- and short-term climatic change s. However, I doubt short-term changes have played a major role in allowing species ranges to expand in areas not influenced by sea-level change. The amount of climate change every 100 Ky is considerable, and, if extinction occurs, may operate over too short a period of time for recolonization given the difficulty of moving between drainages. Further, when sea levels are lowest, climate is typically driest in the tropics (Williams, 1984), possibly countering opportunities for dispersal there.

Interpretation of relationships between inland drainages (LEB, BULL, and MDB) and surrounding regions is difficult. However, the following seems clear, fishes have been exchanged between the Northern Province and LEB, LEB and BULL, MDB and BULL and/or LEB, MDB and SWV, MDB and one or more of FITZ, SEQ, and NEN, and possibly also between SEN and SEV (Fig. 16). Some species are common to many regions, some are shared by only a few, and some by only two. Few species have common range boundaries. Why m ight some species have been exchanged and others not (assuming the cause of exchange did not discriminate between species )? It seems likely the relationship between populations of widespread species will be complex and difficult to unravel, due to possi ble faunal exchange from several regions and different and possibly multiple times and/or directions. An additional difficulty in interpreting patterns of species occurrence among these regions is the question of how and when fish crossed a drainage divi de. Among situations where such crossings are indicated, the drainage rearrangement between Flinders River and Praire Creek (SGC and LEB) is well defined and dated (Coventry et al., 1985). One other, the location of connection between LEB and BULL (see Discussion) appears intuitively obvious, although undated. To my knowledge, no other locations or mechanisms by which other connections might have occurred are obvious. Based on differences between likelihood of movement between adjacent versus non-adjacent drainages, I predict when a species occurs on both sides of non-adjacent divides that populations on each side will be more similar to those in adjacent drainages than to each other.

Excepting parts of northern Australia, coexisting congeners are unusual, except for Galaxiidae, Pseudomugilidae, Percichthyidae, Philypnodon, and Hypseleotris. Craterocephalus spp. often are sympatric, but the congeners are of distinct lineages (Crowley, 1990), with species within a lineage never sympatric except perhaps in one habitat at Dalhousie Springs (LEB; Unmack, 1995). This suggests when congeners come into contact, sympatry is temporary, ending in extinction of one taxon and/or hybridization after mixing, another potential hypothesis for testing.

Today, biogeographic studies lacking phylogenetic data are uncommon. Furthermore, the techniques applied here are not commonly used. Few studies of this nature have combined analytical techniques (however, see Hugueny & Lévêque, 1994). Most examine data using only one technique; either clustering (Warren et al., 1991), ordination, (Reshetnikov & Shakirova, 1993) or parsimony analysis (Watanabe, 1998). Clearly, a minimum combination of clustering and ordination should be used as they to some extent complement each otherís weaknesses. Phylogenetic data are lacking for fishes over most of the world, and this situation likely will continue for some time. I suggest the present type of study based on distributional and geologic data is a useful precursor to phylogenetic studies, as it provides testable hypotheses that would otherwise be unavailable until far into the future.