Crisp & Chandler, Paraphyletic species 
819 
their approach is likely to encounter serious sanapling problems: if the terminals 
comprising a data set are too sparse a sample of the variation within the study 
group, then an incorrect topology may be found because divergent change may 
confound estimates of homology. This is becoming evident in studies using molecular 
data (Melnick et al. 1993; Smith 1994a), as well as fossils (Donoghue et al. 1989: 
444 - 449 ), By analogy, a single individual may be an inadequate sample of the 
character variation within a species or higher taxon. Clearly cladistic analysis cannot, 
by the nature of its assumptions and limitations, reconstruct all the historical events 
affecting populations, such as reticulation or isolation of a lineage in which no 
detectable apomorphy has evolved. However, it should retrieve the major divergence 
events as well supported clades, and on this basis we shall proceed. 
Daviesia ulicifolia 
Recently we have attempted to resolve the complex species Daviesia ulicifolia 
(Fabaceae: Mirbelieae). As currently circumscribed, this taxon extends from Cape 
York Peninsula in far north Queensland (16°S, 145°E) south to Tasmania (43°S, 147°E) 
and west to the Great Victoria Desert (30°S, 124°E). This is a vast range: 27° in 
latitude and 23° in longitude; however, the distribution is 'L'-shaped and only covers 
about 20% of the Australian continent. Additionally, in south-eastern mainland 
Australia, it extends from sea-level to tree-line at 1800 m altitude, where continuous 
snow cover is experienced in winter months. Not surprisingly, this is a polytypic 
species showing several morphological forms. Bentham (1864: 81) named four forms 
(under the illegitimate name D. ulicina), but neither he nor his successors have 
produced a satisfactory classification of the species. Our treatment used mainly 
morphometric characters of the leaves, stems, inflorescence and flowers to identify 
phenetic clusters that might be recognised as taxa. Although several morphologic- 
geographic clusters were evident, most of these intergraded in the regions of overlap, 
and we have treated them either as subspecies or informal forms (Table 1; Chandler 
& Crisp, in prep.). Environmental variables such as soil texture, nutrients and climatic 
parameters correlate with the clusters but these too overlap between forms. The 
only form which we are segregating at species level is the 'Yorke' form, which is 
autapomorphic and appears more closely related to D. arthropoda than to D. ulicifolia 
(Fig. 2; Chandler & Crisp, in prep.). 
We have made a cladistic analysis of the forms of D. ulicifolia and related species 
which share with it a distinctive kind of calyx with equal teeth (D. acicularis, 
D. areuaria, D. arthropoda and D. microcarpa: Pate et al. 1989; Crisp 1995a). At this low 
taxonomic level, most of the available characters are morphometric in nature, and 
we used the coding method of Thiele (1993a). The few qualitative characters tend to 
be autapomorphies for the recognised species (Table 1), for example: toothed and 
revolute leaf margins (D. acicularis); midrib more prominent abaxially (D. areuaria); 
leaves angular-terete, stems lax and unbranched (D. microcarpa). Both D. arthropoda 
and D. ulicifolia lack autapomorphies and should be considered a priori metaspecies. 
For an outgroup we used D. ivyattiana, which appears to be closely related to the 
D. ulicifolia group (Pate et al. 1989). Tables 2 and 3 show the character list and data 
matrix respectively. 
We used the 'branch and bound' algorithm in PAUP (Swofford 1990) to find a single 
most parsimonious tree of 396 steps (Fig. 2). The data set shows significant cladistic 
structure (PTP < 0.01: Faith & Cranston 1991). However, the low bootstrap values on 
most nodes indicate a weak hierarchical pattern in the characters used. Little 
phylogenetic structure is expected at the level of diverging geographic forms, because 
they are unlikely to be fully differentiated lineages due to reticulation or gene flow, 
even if this occurs at a reduced level compared with undifferentiated populations 
