Studies in the Arundinelleae (Gramineae) VIII. The Phylogeny — A Hypothesis J.B. Phipps Department of Botany, The University of Western Ontario, London, Canada Contents Page Abstract 477 1. Introduction 477 2. The nature and interpretation of phylogenetic evidence 478 (i) Phylogenies hypothetical 478 (ii) The interpretation of phylogenetic evidence 478 (iii) Conservative characters 478 (iv) Selection of advanced and primitive character states 479 The (v) character states used 479 (vi) Number of characters used 480 The 3. representation of a phylogeny 480 (i) Principles for arranging cladograms 481 classifications 4. Phyletic 481 Criteria for this 5. accepted paper 481 6. The phylogeny deduced 482 Individual (i) genera 482 Discussion of within (ii) genera groups 485 (iii) The tribe as a whole 487 7. Phyletic conclusions 488 (i) Geographical considerations 488 (ii) Parallelism 489 (iii) Evolutionary reversals 490 (iv) Postulation of a primitive or proto-Arundinellean 490 A (v) phyletic classification 491 8. Summary 491 Acknowledgements 492 References 492 Abstract This study considers the 163 species accepted as belonging to the tribe Arundinelleae (Gramineae) and into A is arranges them a putative cladogram. discussion ofthe rationale presented, 38 characters are studied for advanced versus primitive states, advancement indices calculated, and trends of variation discussed. The six major groups ofPhipps (1966b) are maintained. The phylogeny conforms excellently with the of the continental drift it for the geographical aspects hypothesis though requires a greater age Angiosperms than is generally held to be the case. I. Introduction The the with the tribe Arundinelleae present paper, eighth in a series dealing variationin amine derive a of a (Gr ae), is an attempt to tentative phylogeny group quite intensively 21 478 BLUMEA VOL. XV, No. 2, 1967 studied morphologically (Hubbard 1936, Keng 1936, Bor 1955, Conert 1957, Phipps Lubke & 1964, 1966a and b, 1967a, Correia, Phipps, 1967) but lacking any known fossil record. The classification of be considered be the basis for this Phipps (1966b) can to paper. taxonomic lack In that study some structure emerges, but not much, there being a of numerical The rationale ofthat detailedhierarchisation and no analysis. paper was mainly into six single linkage augmented by a proposed splitting of the tribe parts, though utilised in conclusions. rigorous numerical techniques were not an attempt to justify Phyletic considerations (the formationof groups A-F) were permitted to split Loudetiopsis sensu Conert even though phenetic evidence might have suggested the retention of Diandrostachya in Loudetiopsis. This paper is a phyletic explanation of the classification of Phipps 1966b as that paper has been amended by Phipps 1967a and Correia, Lubke & Phipps, 1967. It includes a discussion of the rationaleused, the characters used, the phylogeny derived, and an infer- ence on the primitive Arundinellean or proto-Arundinellean. 2. THE NATURE AND INTERPRETATION OF PHYLOGENETIC EVIDENCE (i) Phylogenies hypothetical Phylogenies must inevitably be no more than hypotheses, so weak is the reliably evidence the interpretable and, as a result, correspondingly strong speculation, except The evidence in rare extremely well-documented cases. for a phylogeny is, of course, the which straightforward: it is same data (characters) from a phenetic taxonomy is drawn. The data will preferably be drawn from all members (i.e. fossil and extent) of the under consideration and the most reliable group presumably phylogenies are of those based on a good fossil record. In such cases reliability depends on the number fossilised individuals and on the number of characters which are preserved. In the ideal form case the individuals a (furcate) continuum in taxonomic (phenetic) hyperspace and geological time. (ii) The interpretation of phylogenetic evidence these 'well-documented' there need be little about the In situations, argument accuracy of the main lines of phylogeny produced. However, not infrequently there are gaps in the known fossil record and it is here that a combination ofintuitionand speculation (less kindly, 'guesswork') together with Simpson's 'aesthetic sense' (1961, p. 4) help with dotted the that the us to extrapolate our phylogenetic line, through gaps are empty regions of taxonomic hyperspace. Are of the uncharted than there any more reliable means navigating empty areas those mentionedin the previous sentence? It seems that, subject to degrees of probability difficult be to calculate, there may be. The fundamental problem to solved since attempts at phylogenies have become fashionable involves defining the particular distinction between the more primitive (i.e. cladistically basal) and the more advanced (cladistically acropetal) states of a given character, and this is discussed further on. Meanwhile certain other factors must be discussed. (iii) Conservative characters In a diverse characters the sufficiently group many significant (= obvious) vary. In old days of phylogeny the more 'fundamental' (because 'conservative') characters (in our terms: conservative character 'states') would be used to indicate the main delineations J. B. PHIPPS: Studies in the Arundinelleae (Gramineae) VIII. Phylogeny 479 ofthecladogram butit is now understood that as a premiss suchan approach is unworkable for much parallelism exists in the phylogeny of any sufficiently diverse group (i.e. most sufficiently large groups) - an observation empirically attestable if one cares to juggle with cladistic eliminate characters re-arrangements to it. But though conservative are those displaying little or no parallelism and these characters may be known from the of in inspection any adequately annotatedcladogram, nonetheless, a sufficiently complex situation (as is common) various alternative sets of character states could be made cladistically basal and consequently conservative. Accordingly, conservative characters, should have tended be used for though they emerge as a conclusion, to as a premise, there unfortunately, in practice, are rarely cogent reasons adduced for their choice therefore the of of that which avoid circularity. It is point view the present writer nothing characters whatsoever may be mentioned of conservative states except as conclusions. (iv) Selection of advanced and primitive character states and There remains the problem of deciding on the particular advanced primitive of each character. there is least universal of states Apparently no good or at way doing this. The principle criterion is the contention that more 'specialised' (i.e. more complex) this character states are more advanced than less 'specialised' ones. However, assumption lead absurd results. it be allowed may to apparently (at least, unlikely) Consequently, must that acomplex state can sometimes proceed to a more simplified one, an escape fortunately justified by the example of certain well-documentedphylogenies. Other criteria com- monly used are that vestigial characters are advanced (really a special case of complex state proceeding to a more simplified one) and that embryonic or early ontogenetic character tend be 'conservative' be treated with states to (a criterion to great caution). In has correlation character addition Sporne (i960) demonstrated a fairly strong among state sets in Angiosperm families, the one set being primitive, the other advanced, though there is no criterion inherent in his statistical investigation for saying which is which. There has been little follow very up on Sporne's approach. Wagner (condensed in Benson, 1962), using in-group—out-group comparisons, criteria this time based the taxonomic gives yet further for probable primitiveness, on distribution of character states, e.g.: 1) that which is most primitive exists in all taxa of the next lower rank and all nearly related taxa of the same rank; 8) that which is most advanced exists in the minority of the taxa of the next lower rank the rank. and in none of the taxa of same Wagner in his duplicated class-sheet lists 6 intermediate situations on a graded series. Wagner's approach has been used by litis (1959), White (1963), and various other workers. such when used for the likelihood of However, criteria as Wagner's, actually any rather given character state being primitive or advanced, may seem alarming, in principle, have defended basis and to use upon a new group. They never been on a logical seem to be intuitive. If they someday prove to be correct as generalisations they will be very important criteria. However, to be proved correct, they must be shown to be general which have used them and this has in phylogenetic cases not as premises, investigation yet to be made. (v) The character states used well described The literature indicates that the Arundinelleae are quite morphologically. data are available the and floras and Morphological in type descriptions in very many 480 BLUMEA VOL XV, No. 2, 1967 each nay be sought using the references given after species in Phipps (1966b) and, for and very recently described taxa, in Correia, Lubke Phipps (1967). A general description variations in Arundinellean characteristics be found in the discussion in of may Phipps and the table utilise 1964 (pp. 88—92) in (op. cit., opposite p. 120). However, we cannot all known characteristics for deducing our phylogeny since for a number we have no opinion whatsoever as to which state might be advanced and which primitive. Table 1 illustrates the characters actually selected. (vi)