a century of progress in grass systematics

IntroductionAlthough the c. 10,000 species in the Poaceae places itas the fifth largest flowering plant family, a number ofbiological features raise it to a unique position. Thefamily has contributed crop species that provideabout 80% of the annual global food (FAOSTAT1999). In fact, four of the top crops that feed theworld are cereal crops: wheat, rice, corn, and barley.This economic significance is matched by anecological dominance as grasses cover about one fifthof the earth (Shantz 1954). Grasslands are signatureof the Poaceae. Because of this economic andecological importance, grasses have attractedconsiderable attention in the biological andagricultural sciences. The past century showedenormous amounts of research conducted on variousbiological aspects of the family, including systematics,genetics, cytogenetics, breeding, physiology, anatomy,developmental biology, chemistry, chromosomestructure, and, more recently, whole plastid andnuclear genome sequences. Results from thesestudies have provided a wealth of information on thebiology of the family, which have become and remain

valuable resources used in furthering ourunderstanding of Poaceae systematics and in assessingthe mode and tempo of grass evolution.

Development of our views of grass systematicsfollowed major trends in advancement in the sciencesand engineering. Major discoveries come in spurtsand are usually followed by periods of relativestagnation. Optical development associated withmicroscopes provided a closer look at variousanatomical features; advancement in chemicaltechniques revealed information on a number ofmolecules such as flavonoids, proteins and alkaloids;development of computers and various analyticalsoftware programs enhanced the capability ofhandling large data sets and facilitated large-scaleanalyses of available grass data; and finally, thecurrent advancement in molecular biology andbiotechnology provided us with the tools to look atinformation derived from genes and genomes.

Parallel to, or as a consequence of, these majordevelopments, the field of grass systematics underwenta series of refinements that individually or collectivelycorresponded to the progressive stages cited above.

355KEW BULLETIN 62: 355373 (2007)

A century of progress in grass systematics

Khidir W. Hilu1

Summary. This paper presents an overview of progress in grass systematics with a focus on the past century andassesses its current status and future outlook. In concert with systematic biology, progress in grass systematics hasgone through some leaps caused by the introduction of new approaches or emphasis on existing ones.Chromosome cytology, anatomy and chemistry provided useful information, but major recent contributionshave come from advances in bioinformatics and molecular biology. Consequently, grass systematics has movedfrom an initial intuitive classification and phylogenetics to one incorporating analytical phenetic approaches,and culminating in the current stage of analytic phylogeny. As a result, a refined picture of grass phylogeny isemerging with good resolution at the base, but the tree lacks robustness in some places such as the monophylyof the BEP subfamilies and the relationships within the PACCAD clade. Systematic structure of a number ofsubfamilies is better understood now, but further studies are needed. With the rapid advancement in molecularsystematic and bioinformatic tools, and in conjunction with a wealth of literature available on structuralcharacters, a more refined picture of grass taxonomy and evolution is expected. However, caution needs to beexercised in our interpretations to avoid hasty decisions that can translate into regress rather than progress. Thisis an exciting time in the history of grass systematics and, undoubtedly, is a period of collaborative rather thanindividual effort.

Key words. Poaceae, grasses, systematics, phylogenetics, history, evolution.

Accepted for publication November 2006.1 Department of Biological Sciences, Virginia Tech, Blacksburg, VA 24061, U.S.A. Phone: 1-540 231 6407; Fax: 1-540 552 9307; e-mail: [email protected]

The Board of Trustees of the Royal Botanic Gardens, Kew, 2007

These stages of progress in grass systematics, althoughnot necessary sharply distinct, are in general quiterecognizable and can be divided into two majorperiods, which in turn can be further divided into sub-periods. The two main periods are the PredictivePeriod and the Analytical Period. These two periodsdiffer in the means by which hypotheses on grasssystematics are put forth. In the Predictive Period,taxonomic hypotheses were based on personalperceptions using available knowledge in the field andpersonal conviction of the systematist. In the AnalyticalPeriod, hypotheses are based on analyses of availabledata with preset mathematical formulae that haveeventually taken advantage of computing systems. Thispaper will address grass systematics in these twoperiods, summarize the overall progress to date, assessthe areas that need further work, and highlightpossible future trends and needs in grass systematics.

Predictive Period in Grass SystematicsThis period began with the first publication on theclassification of grasses by Adanson (1763). Theperiod included both the classification of grasses intosubfamilies and tribes (Predictive Taxonomy) as wellas the assessment of evolutionary patterns for thetaxonomic units (Predictive Phylogeny). Althoughthe two approaches overlap completely in time, itwould be best to address them separately.

Predictive TaxonomyHilu & Wright (1982) recognized two stages in thisperiod, an initial stage that relied solely onmorphological characters, and a second stage thatgradually incorporated microscopic structures andbiochemical features. The initial stage started with

Adansons (1763) division of the grass family intoseveral sections. Outstanding contributions duringthat period included the allocation by Robert Brown(1814) of grass genera into the tribes Panicaceae andPoaceae (probably equivalent to what was laterrecognized as subfamilies Panicoideae and thetraditional Festucoideae) based on detailedunderstanding of spikelet structure. This system wasfollowed to a certain degree by Bentham (1881), andbecame the basis of the classification of Hitchcockand Chase (1950) of the North American grasses intothe subfamilies Festucoideae and Panicoideae. Betweenthe 1673 and the 1950 publications, a number ofcontributions were made in which the number ofsubfamilies ranged between two and three(summarized in Hilu & Wright 1982 and GPWG2001). The recognition of the major subgroups wasbased largely on type of inflorescence, spikeletnumber and floret disarticulation.

The emergence of additional characters in grasssystematics had an early start with the recognition inthe nineteenth century of useful patterns in starchgrain (see Hilu & Wright 1982 and GPWG 2001). Thisinformation attracted the attention to the presence ofimportant traits besides morphology and encouragedfurther work on these same traits or others by a newgeneration of researchers, ultimately leading tosignificant advances in grass systematics. These newlines of research included (but were not necessarilylimited to) microscopic features, chromosomecytology, leaf and stem anatomy, embryology, seedlingtype, photosynthetic pathways, seed storage protein,flavonoid chemistry (see literature review in Hilu &Wright 1982; Clayton & Renvoize 1986; GPWG 2001).The Predictive Period began with what seems to besimplistic view of grass classification and moved into

356 KEW BULLETIN VOL. 62(3)


Table 1. Selected taxonomic treatments of the Poaceae at higher taxonomic levels to represent the overall history and thevarious periods discussed here. Brown (1814) classification of grasses into two tribes roughly corresponds to later treatments atthe subfamily.

Brown 1814 Avdulov 1931 Stebbins & Caro 1982 Clayton & GPWG 2001Crampton (1961) Renvoize 1986

Paniceae Poatae Bambusoideae Bambusoideae Bambusoideae AnomochlooideaePoaceae Sacchariferae Oryzoideae Streptochaetoideae Oryzoideae Pharoideae

Arundinoideae Anomochlooideae Pooideae PuelioideaeFestucoideae Olyroideae Chloridoideae BambusoideaeEragrostoideae Centhostecoideae Panicoideae Ehrhartoideae

Oryzoideae Arundinoideae PooideaeEhrhartoideae Centothecoideae AristidoideaePhragmitoideae ChloridoideaeFestucoideae CentothecoideaeEragrostoideae PanicoideaeAristidoideae DanthonioideaePanicoideae ArundinoideaeMicrairoideae

the impressive effort of exploring potentialinformative characters in the Poaceae. Thesecontributions clearly underscored the artificiality ofthe two or three subfamily systems and pointed outthe need for substantial taxonomic revisions. As aconsequence, new subfamilies were recognized (Table1) with the number varying from three (Avdulov1931) to 13 (Caro 1982). The highlight of the periodis the instability of grass classification as reflected byinconsistencies in the trend of subfamily recognition(Fig. 1). The obvious reason for the inconsistencies isthe subjective differential weighing of characters inthe process of subfamily circumscription based on theindividual judgments on taxonomic characters andsubfamily criterion.

Predictive PhylogeneticsConcepts on grass evolution emerged late in the 19thcentury (Celakovsky 1889; Goebel 1895; Schuster1910), reflecting keen observations. It was suggestedthat Streptochaeta or a grass with similar spikeletstructure is the most ancestral in the family. Thoseauthors took the first crucial steps in evaluatingcharacter polarity and using it in predicting potentialphylogenetic trends. The theme was exploited insubsequent predictive phylogenetic assessments, suchas those of Stebbins (1956) and Tateoka (1957), withsome of those predictions being confirmed by recentmolecular phylogenetic studies. I will discuss fivepapers from that period to illustrate the phylogeneticthinking of grass systematists at this stage andcompare their predictions with our currentunderstanding of grass phylogeny.

Roshevits (1937) provided what he termed asystem of origin of grasses based on the two-subfamily system (Poatae and Sacchariferae) of Avdulov(1930). Within the two subfamilies, three serieswere recognized for the Poatae (Series Bambusiformes,Phragmitiformes, and Festuciformes) and two for theSacchariferae (Paniciformes and Eragrostiformes). Hedepicted the Poatae emerging from an unknownprogenitors of grasses with the Bambusiformes at itsbase, and the Sacchariferae to be derived from thePoatae at some point. Roshevits (1937) placedStreptochaeteae at the base the Bambusiformes, andconsequently of the whole grass family. FromStreptochaeteae, he envisioned the divergence of thetribe Bambuseae and subsequent bifurcation into thePhareae line and the ancestral type of thePhragmitiformes series. Within the Phragmitiformes, hedrew a line from the Arundineae to the ancestral typesof the Festuciformes. For the subfamily Sacchariferae,he presented an ancestral type at the base, giving riseto the two series Paniciformes and Eragrostiformes.Roshevits also presented detailed branching patternsfor the respective tribes of each of the five series.Despite the problems in the classification system, thepresentation of a predictive grass phylogeny in abranching form, the placement of the Streptochaeteaeat the very base of the family and the Phareae near thebase are important contributions.

Stebbins (1956) presented an illustration of hisperspectives of evolutionary patterns in the Poaceae(Fig. 2A). The outline is not in the form of aphylogenetic tree but rather a view of a tree canopywith an unknown ancestral taxon in the center and the

A CENTURY OF PROGRESS IN GRASS SYSTEMATICS 357


Fig. 1. Trends in Poaceae subfamily recognition since Brown (1814) treatment of the family.



Fig. 2. Patterns of evolutionary relationships among subfamilies and tribes of Poaceae presented after Stebbins (1956). G. L. Stebbinsillustrated the relative differentiation of the grass groups around an ancestral complex (noted as an asterisk) in (A), or (Stebbins 1982) aslines radiating from an ancestral complex (B). In the latter case, measurements of branch lengths are superimposed here on a re-drawing of his original diagram.

A

B

various grass lineages placed in distances thatcorrespond to their degree of specialization. Stebbinsavoidance of depicting branching patterns for thegrass tree was likely driven by caution. He notablyplaced Streptochaeta very close to the root, affirmingprevious concepts. The tribes Arundineae, Danthonieae,Centotheceae, and Bambuseae were also very close to theroot. Pharus, one of the currently recognized earlydiverging genera, was placed in a separate island closeto the Oryzeae but relatively distant from the root. Hepredicated ancestral taxa for some of the subfamilies:Arundinelleae for the Panicoideae; the genera Melica,Glyceria and Schizachne (Meliceae) for the Festucoideae(Pooideae); Uniola for the Eragrostoideae (Chloridoideae);and Streptochaeta for the Bambusoideae. Some of theserelationships are confirmed now, such as the nearbasal position of Uniola in the Chloridoideae (Hilu &Alice 2001) and the early divergence of the Meliceaeand Nardus in the Pooideae (Soreng & Davis 1998; Hiluet al. 1999; Dring et al. this issue). Stebbins (1956)placed Aristida close to the Chloridoideae.

In a subsequent publication, Stebbins (1982)presented his predictive evolutionary concept of grassesin another form with six subfamilies radiating from theunknown ancestor (Fig. 2B). He based his hypothesis

on gross morphological and microscopic charactersand noted character polarity for several traits, includinghabit. The distance from the ancestral group andrelative position depicted degrees of evolutionaryrelationships. I have measured the distances to quantifyStebbins thinking of the degree of evolutionarydifferentiation (Fig. 2B). The Bambusoideae (withStreptochaeta and Bambusa close to the base) has theshortest distance from the ancestral group (3.8 cm),followed by the Pooideae (5.8) and Oryzoideae (5.9). Theevolutionary distances assumed a relative leap for theArundinoideae, Eragrostoideae and Panicoideae,considering their distances at 7.7 cm, 9.0 cm and 12.2cm from the ancestors, respectively. The short distanceof the Bambusoideae from the ancestral group may havebeen influenced by the inclusion of Streptochaeta in thesubfamily. The overall system reflects to some degreethe current thinking in grass phylogenetics byunderscoring close affinities among todays PACCADclade members and their relative distance from theroot of the family, and highlighting some affinitiesbetween Oryzoideae (Ehrhartoideae) and Pooideae.

A year after the publication of Stebbins (1956),Tateoka (1957) presented his views on this subject(Fig. 3). Tateoka (1957) pointed out the importance



Fig. 3. A predictive phylogeny for the Poaceae proposed by Tateoka (1957). Arrows are superimposed on his original lines connectingthe grass subfamilies.

of various features such as chromosome cytology, leafanatomy, embryo structure, seedling type, root hairs,and starch grains in grass systematics. However,among these features, Tateoka consideredchromosome cytology and leaf structure to beparticularly important in grass systematics and usedthem along with morphology and geographicdistribution to present his taxonomic andphylogenetic schemes. Tateoka presented hisperception of divergence patterns in the Poaceae withsolid lines, although the starting points of the lineslinking subfamilies were not connected to aparticular taxon and the divergence patterns withinsubfamilies was quite vague (Fig. 3). Those branchesare highlighted here in arrows. He proposed anunknown taxon at the base of his Pharoideae as aprogenitor for the Poaceae. Impressively, he placedAnomochloeae and Streptochaeteae as the first diverginggrasses after that ancestor, a position that has beenrecently confirmed (Clark et al. 1995; Hilu et al.1999). His Pharoideae included eight tribes of what aretraditionally Bambusoideae and Oryzoideae, with theOryzeae branching off among the bambusoids distantfrom the base. The Pooideae was placed at the top ofthe grass phylogeny, far removed from hisbambusoid-oryzoid group. The Panicoideae wereplaced adjacent to the Pharoideae, as were the

Arundinoideae (Arundoideae of Tateoka, Fig. 3).However, no solid line leading to the Panicoideae wasdrawn, possibly implying uncertainty about its origin.The Arundinoideae had Arundinelleae at its base, withthe latter adjacent to the borders of neighboringPanicoideae (Fig. 3). This placement of theArundinelleae signals a conjecture of potential closeaffinity of the tribe to the Panicoideae where it is nowplaced. From his Arundinoideae, Tateoka derived thePooideae from near Lygeeae and Nardeae, two tribes thatare now placed in the Pooideae where they occupy anear basal position in the subfamily in currentphylogenetic hypotheses (Soreng & Davis 1998; Hiluet al. 1999; Dring et al. this issue). The Eragrostoideaeappear to be derived from near the Arundineae; thisposition has been contemplated in subsequentstudies and our work (Hilu et al. 1999).

Tsvelev (1976) proposed evolutionary relationshipsamong grass tribes found in the USSR. Therelationship is depicted in a circular form withsubfamily placement reflecting relative close affinities,and tribe positions within those subfamilies placed ata relative distance from a central ancestral group. Heasserted that the theme of his evolutionaryperspective followed that of Stebbins (1956). Tsvelev(1976) stated that it is impossible to derive all tribesand subfamilies of grasses from any single existing



Fig. 4. A circular representation of subfamily and tribal phylogeny in the Poaceae put forth by Tsvelev (1976) for the genera of theUSSR. Distances from the center represent degree of differentiation, and subfamily positions are indicative of measures of relatedness toeach other.

tribe. Therefore, it will be more correct to display theaffinity relationships between the grass tribes weaccepted, not in the form of an ancestral tree, but inthe form of a cross section of this tree at this stage, asdone by Stebbins (1956). Tsvelevs note of cautionabout the depiction of grass phylogeny may have beenbased on emphasis of extinction of some ancestrallinks in the family. The three major groups(subfamilies) arundinoids (including the Aristideae),chloridoids and panicoids were placed in an adjacentposition, leaving a neighboring bambusoids, oryzoidsand festucoids (Festucoideae, currently named Pooideae)(Fig. 4). The placement of the Bambusoideae next tothe Panicoideae was mirrored in Stebbins (1982)evolutionary hypothesis. The proximal positions canbe construed as a primal presentation of what laterwas recognized as the PACCAD and BEP clades(Clark et al. 1995). Tribal positions of interest includethe placement of the Arundinelleae closer to the centerof the diagram within the Panicoideae and the basalposition of the Pappophoreae in his Eragrostoideae,implying overall relative ancestral status. Thesephylogenetic positions have recently been confirmed(Hilu et al. 1999; Hilu & Alice 2001; GPWG 2001). It israther intriguing to see that Tsvelev presented all thetribes in his diagram as distinct entities except for his

Poeae-Aveneae-Phleeae-Phalaroideae group. Recentmolecular phylogenetic studies have clearlydemonstrated the difficulties in segregating membersof the traditional Aveneae and Poeae into twomonophyletic lineages (Dring et al., this issue).

Among the most extensive and detailedpresentations of evolutionary relationships are thosepresented by Clayton & Renvoize (1986) in GeneraGraminum. Their perspective of subfamilyrelationships underscores the importance of theevolution of photosynthetic pathways and theirassociated anatomical and ecological features. Clayton& Renvoizes (1986) contribution to predictiveevolutionary relationships was immense at the tribaland genus levels where hypotheses on patterns ofdivergence were presented for each subfamily andsome individual tribes. These predictive evolutionaryrelationships have been utilized as starting hypothesesfor a vast number of subsequent studies, includingpresent day phylogenetic reconstructions. Theproposed evolutionary relationships were furthersupported by information in their text, furthering thesignificance of the publication. In numerous cases,some of the genera, particularly the large ones, aredepicted as ancestors to more than two lineages (e.g.Tridens, Eragrostis and Leptochloa in Fig. 5). However,



Fig. 5. One of the diagrams illustrating Clayton and Renvoizes (1986) views of evolutionary relationships among grass genera andtribes.

polyphyly and paraphyly of some of these genera havebeen demonstrated (Hilu & Alice 2001), and thisapproach of presenting evolutionary relationship inpredictive phylogenetics should not be construed as adrawback but rather a strength in the predictions.

Analytical PeriodWith these assessments of grass systematics underway,analytical approaches of data sets of various sizes andcomposition began to surface. The development incomputer technology coupled with advances incomputational biology, especially those directed towardsystematic biology, were major factors underlying theemergence and subsequent success of the AnalyticalPeriod. This period will be further divided into twotypes: Phenetic Analyses and Phylogenetic Approaches.

Phenetic AnalysesThe Phenetic Analyses period began with the adventof Numerical Taxonomy and its application insystematic biology. The field was ideal for grasssystematics with the availability of large amounts ofinformation in the literature on the vast majority ofgrass genera. Numerical taxonomy, besides itsmanipulation of large data sets (taxa andcharacters), helps in reducing, but not eliminating,subjectivity in handling and assessing information, apractice that took us a step closer to naturaltaxonomic treatments. Although the handling oflarge data sets is an accurate assertion, subjectivitycan be introduced through the choice of characters,their scoring and the delimitation of groups, i.e. theplacement of the phenon lines that demark thesplitting of clusters. The impact of these variables



Fig. 6. A phenogram from the Clifford et al. (1969) cluster analysis study of Poaceae. The re-drawn diagram accurately reflects thedistances noted in the original publication.

was evident in the numerical phenetic studiesconducted on the Poaceae. I will focus here on fivestudies that aimed at assessing grass taxonomy at thesubfamily level.

H. T. Clifford pioneered the field of numericalanalyses in the Poaceae (Clifford 1965; Clifford &Goodall 1967; Clifford et al. 1969; Watson et al. 1985).I will summarize here two of these studies. Clifford etal. (1969) scored 50 characters for 92 grass generaand analyzed the data set with MULTCLAS andMULTBET for cluster analysis and GOWER forprincipal coordinate analysis (PCO). Theyrepresented each genus by one species, a practicethat may not account for variability at theintrageneric level. The highlight of the analysis wasthe presence of two major clusters of genera (Fig. 6),one comprising the pooid and bambusoid grasses astwo distinct subclusters. The second cluster includesthe andropogonoid grasses in a subcluster linked tothe eragrostoids and two closely associatedsubclusters representing the panicoid andphragmitoid groups of grasses. Their PCO analysissegregated the bambusoids from the remaininggroups, but again separated the panicoids from theandropogonoid grasses. The emergence of two majorclusters is an important contribution as they reflectwhat was later defined as BEP and PACCAD clades.Worth noting here is the low degree of correlationbetween the pooids and bambusoids (Fig. 6), a findingthat underscores the looseness of the latter group.

Based on the five groups emerged in the clusteranalysis and the distribution of the taxa in the PCO,Watson and Clifford concluded that grasses could bedivided into at least five groups of homogenousbambusoids, pooids and andropogonoids, and therelatively heterogeneous eragrostoid and panicoid-phragmitoid groups. They noted that due to the lackof discreteness among the groups and pattern ofdistribution of the genera in the PCO, it is notpossible to recognize some of the groups at thesubfamily level.

Watson, Clifford and Dallwitz (Watson et al. 1985)conducted another numerical analysis on the Poaceaebut this time with a larger data set. They analyzed adata set of 720 genera and 85 morphological andmicroscopic characters with the SAHN and MACINFprograms. They recovered similar clustering patternswith both programs, but in the case of using SAHN,they limited the taxa to near-comprehensivelyrecorded taxa and noted that they had tolaboriously weigh all the information they had tohand. The summary dendrogram in Watson et al.(1985) shows two major clusters, one representingthe subfamily Panicoideae, but this time all itsmembers (recognized as supertribes Panicanae andAndropogonanae) clustered together. The secondmajor grouping included the subfamilies Pooideae,Bambusoideae, Chloridoideae, and Arundinoideae (Fig.7). The grouping into five well-defined clustersrepresenting these subfamilies was an improvement



Fig. 7. A redraw of the summary dendrogram from Watson et al. (1985) cluster analysis study showing grass subfamilies and majortribes. The arrows highlight the two major clusters and the circle points out the unusual grouping of the chloridoid and arundinoidgrasses with the pooids and bambusoids.



over the study of Clifford et al. (1969). However, thevery distinct separation of the Panicoideae from therest of the family and the nesting of the Chloridoideaeand Arundinoideae in the Bambusoideae and Pooideaecluster was later clearly demonstrated to be artificialrelationships in numerous studies as well as currentfindings (see Clark et al. 1995; Hilu et al. 1999;GPWG 2001).

Hilu & Wright (1982) conducted a cluster analysisstudy on 215 grass genera and 85 morphological andmicroscopic characters using the unweighted pair-group method with simple averages (UPGMA). Inthat study, variation in traits within genera wasaccommodated by coding the presence and absenceof each character as separate bistate characters (fordetails see Hilu & Wright 1982). Based on this typeof scoring, the number of characters used in theanalyses totaled 220. All characters were weightedequally. Eight clusters were evident based on totalcharacters. The clusters corresponded to thesubfamilies Pooideae, Oryzoideae, Bambusoideae,Eragrostoideae (Chloridoideae), Arundinoideae,Centostecoideae (Centothecoideae), and a cluster that

encompasses Lygeum, Nardus and Diarrhena (Fig. 8).The latter three genera were proposed as a potentialnucleus of a new subfamily, Nardoideae. Currentmolecular-based trees are generally congruent withthe overall topology of that phenogram. Hilu andWright noted that Pharus displayed extremeisolation. The genus clustered first with theBambusoideae (Fig. 8). After the Bambusoideae, grasseswere assembled into two clusters: one includedPanicoideae, Chloridoideae, Centothecoideae, andArundinoideae and the second the Oryzoideae and theFestucoideae (Pooideae) plus Lygeum, Nardus andDiarrhena. The clustering, thus, represent Pharus,and what was later dubbed PACCAD (Panicoids,arundinoids, chloridoids, centothecoids, aristoids,danthonoids), and BEP (bambusoids, ehrhartoids,pooids) with the Bambusoideae as a separate cluster(Anomochloa and Streptochaeta were not included inthe study). The position of Lygeum, Nardus andDiarrhena in relation to the Pooideae is also quiteimportant as, at least Lygeum and Nardus, areappearing at the base of the Pooideae in recentphylogenetic studies (Soreng & Davis, 1998; Hilu et

Fig. 8. The overall clustering of grass subfamilies from Hilu and Wright (1982). The diagram is re-drawn with Pharus added to reflect itsposition in the dendrogram. Narrow arrow defines the cluster corresponding to PACCAD clade, whereas broader arrows point to thepositions of Bambusoideae, Festudoideae (Pooideae) and Oryzoideae (Ehrhartoideae) (BEP). The circle points to the position of Lygeum,Nardus and Diarrhena (referred to as Nardoideae) in relation to the Festucoideae (Pooideae).

al. 1999; GPWG 2001). Dring et al. (this issue)recovered a basal clade comprising Lygeum andNardus (Diarrhena was not represented) in thePooideae in a matK-based tree, which they arerecognizing as a nucleus for a tribe Nardeae.Therefore, in the Hilu & Wright (1982)dendrogram, the Pooideae cluster can be construedas core Pooideae (Festucoideae) linked to the remainingpooids Lygeum, Nardus and Diarrhena.

The last phenetic case to be discussed is based onmolecular characters: the grain storage proteinprolamins (see Hilu 2000 for review). Theserepresent a series of studies started in 1988 (Hilu &Esen 1988) on prolamins polypeptide size, andimmunological similarities (as measured by degreeof cross-reactivity using antisera) at the wholefamily level and within individual subfamilies.Immunological similarities in this work weremeasured by the highly sensitive enzyme-linkedimmunosorbent assay (ELISA) method andquantified by an ELISA reader. Prolamins range inmolecular size from about 10 kDa to over 100 kDaand are encoded by a multigene family. The sizevariation is quite linked to the current systematicopinion of the family, with the Oryzoideae andBambusoideae having 10 16 kDa prolamins, the

Chloridoideae, Panicoideae, Arundinoideae (s. l.),Centothecoideae, and Aristida displaying 20 30 kDapolypeptides, and the Pooideae having a wide rangeof polypeptide sizes but primarily around 30 100kDa. These prolamin size groups were the firstmolecular markers to delimit a major grass lineagethat was later named PACC/PACCAD. The studiesalso pointed out similarities between theBambusoideae and Oryzoideae in prolamins, which waslatter confirmed by DNA sequences of the 10kDagenes in representatives of the two subfamilies(Hilu & Sharova 1998). These size similarities werefurther examined by ELISA to assess structuralsimilarities among prolamins. An UPGMA analysisof immunological data from representatives ofthese grass subfamilies substantiated the prolaminsize similarities and their derived relationships. Thegrouping showed the bambusoids to be verydistinct, followed by the oryzoids, the pooids andthen a cluster representing members of thePACCAD clade with Aristida being least similar (Fig.9). The tribe Aristideae was shown by matK data(Hilu et al. 1999) as sister to the rest of thePACCAD clade. Thus, prolamin size and molecularsimilarities did not resolve a BEP but a well-definedPACCAD.



Fig. 9. An illustration reflecting phenetic relationships among grass subfamilies based on immunological similarities of prolaminsmeasured by the ELISA method; polypeptide sizes are mapped on the phenogram. The diagram summarizes results from a series ofpublications by K. Hilu and A. Esen reviewed in Hilu (2000).

Analytical PhylogeneticsThe 1986 international conference on grassessystematic organized by T. Soderstrom, K. Hilu, M.Barkworth, and C. Campbell and held at theSmithsonian Institute in Washington D.C. and itsproceedings (Soderstrom et al. 1987) represents alandmark in grass systematics. It brought together alarge number of scientists with diverse interests ingrasses and resulted in presentations and a subsequentpublication that highlighted current knowledge andpotential future directions in the field. That volumecontained the first two publications on cladisticanalyses of the Poaceae (Baum 1987; Kellogg &Campbell 1987). Both studies were morphology-based.Kellogg & Campbell (1987) based their analysis on aninitial 390 grass genera using 23 to 33 morphologicaland anatomical characters. They established themonophyly of the Pooideae, Bambusoideae, andPanicoideae, recognized the polyphyly of theArundinoideae, and pointed out the lack of conclusiveevidence for the monophyly of the Chloridoideae. UsingJoinvillea as the outgroup, they illustrated one of seven

most parsimonious trees; the tree showed Pooideae plusNardus as the first diverging lineage in the Poaceae, withthe Aristideae diverged near the base. Kellogg &Campbell (1987) concluded that tree structuredepends largely on assumption of structure withinsubfamilies. They also pointed out a high proportionof homoplasy as indicated by the low consistencyindex. High degrees of homoplasy, which is notunexpected in morphologies, could greatly impacttree robustness by increasing the proportion of noisecompared with signal.

In contrast, Baums (1987) analysis was based onfewer taxa (21 tribes from North America), a smallernumber (21) of characters, and lacked an outgroup.Baum stressed that his cladistics study was preliminaryand tentative in nature. Baums (1987) phylogenetichypothesis was impressive for an initial attempt at thefamily as well as for being based on morphologicalcharacters. The tree showed the bambusoidArundinarieae at the base, followed by Phareae,Oryzeae, Pooideae, and then a lineage corresponding tocurrent PACCAD (Fig. 10). In the latter group, the



Fig. 10. One of the first two cladistics hypotheses on grass phylogeny re-drawn from Baum (1987). The diagram highlights thephylogenetic positions of major grass groups, with an arrow added to denote the PACCAD clade.

Centotheceae was sister to a monophyletic Panicoideaeplus remaining tribes. The Chloridoideae appearedpolyphyletic and included the Stipeae, Aristideae andArundineae. As in the Kellogg & Campbell (1987)cladistics analysis, homoplasy in structural charactersmight have impacted the analysis, particularly theterminal lineage.

With the advent of molecular phylogenetics, someof the first phylogenetic attempts were based first ongene sequence information with limited taxonsampling and then restriction site analysis from thechloroplast genome. I will discuss first thecontributions of restriction site studies but will addressthose initial sequencing studies with the rest of thecontributions of this approach. Davis & Soreng (1993)carried out the first restriction data study to addresssystematic questions in the Poaceae at the subfamilylevel. The study focused on the Pooideae but includedrepresentatives of other subfamilies. Rooted withJoinvillea, they obtained eight most parsimonioustrees. In the tree they presented (one of the eightmost parsimonious tree), two major clades wererecovered. One clade included members of thePanicoideae, Arundinoideae, Centothecoideae, andChloridoideae, which they referred to as PACC clade, assister to a clade encompassing the bambusoid andoryzoid genera with Nardus and Pharus nested inthem. The second clade resolved the pooid genera. Ina follow-up study, Soreng & Davis (1998)phylogenetically analyzed a data set of 364 chloroplastrestriction sites and 42 structural characters(morphological, anatomical, physiological,chromosomal, and plastid genome structuralfeatures) for 72 grass genera and three outgroupspecies from the Restionaceae, Flagellariaceae andJoinvilleaceae. The study was again focusing on thePooideae and, thus, the sampling was skewed towardthat subfamily. They conducted data and taxapartitioning in that study; I will focus here on theresults of the combined data analysis. The strictconsensus tree of the combined data rooted withRestionaceae, Flagellariaceae and Joinvilleaceae, depicted agrade of Streptochaeta + Anomochloa, Pharus, sister to apolytomy of Oryzoideae, Bambusoideae (in two successiveclades representing woody and herbaceous taxa).Following this grade, the remaining taxa formed apolytomy of Brachyelytrum, remaining Pooideae and aPACCAD clade. Although the backbone of the treelacked robustness, resolving Lygeum and Nardus as firstdiverging in the Pooideae, bringing insight intorelationships among members of the subfamily, andresolving the sister group relationship of Streptochaeta+ Anomochloa to the remaining members of the Poaceaeare important contributions. However, the monophylyStreptochaeta and Anomochloa has been disputed(Clayton & Renvoize 1986; Hilu et al. 1999; Mathews etal. 2000; Zhang 2000).

Restriction sites were soon replaced in molecularsystematics by a more superior molecular approach,namely, nucleotide substitutions and indels (insertiondeletion events). These characters are easy to define,cumulative when compared with restriction site data,and accessible through GenBanks. Hamby & Zimmer(1988) and Doebley et al. (1990) started this field ofwork in the Poaceae utilizing sequences from nuclearribosomal RNA and the plastid rbcL gene sequences,respectively. Small taxon sampling, as the field wastechnically in its infancy, limited their contribution tograss phylogeny. Two studies appeared in 1995addressing phylogenetic relationships based onnucleotide substitutions, one focused on theArundinoideae (Barker et al. 1995), and the other onthe Bambusoideae (Clark et al. 1995) with broadersampling from other grass groups. Barker andcolleagues, using the rbcL gene sequences, constructeda phylogenetic tree showing a paraphyletic Oryzoideae(Zizania appeared sister to an Oryza plus Bambusaclade) as the first diverging Poaceae (Anomochloa,Streptochaeta and Pharus were not included). Thisclade was followed by a split in the family intoPooideae plus a well supported clade that reflected apolyphyletic Arundinoideae (included Chloridoideae)and a centothecoid clade that was sister to thePanicoideae. The polyphyly of the traditionalArundinoideae is the highlight of that study. Asubsequent study (Barker et al. 1999) based onsequences from another plastid gene, rpoC2, resultedin the redefinition of the Arundinoideae and thesegregation of the Danthonioideae. Using sequencesfrom the plastid ndhF gene, Clark et al. (1995)generated a tree in which the monophyly ofStreptochaeta and Anomochloa and its sister grouprelationship to remaining grasses emerged, reflectingthe relationship appeared in the study of Soreng &Davis (1998). However, the novel contribution oftheir study to the backbone of grass phylogeny wasthe emergence of two major clades, one is the PACCand the other is a newly named BOP to encompassthe Bambusoideae, Oryzoideae and Pooideae. Clark et al.(1995) did not cite bootstrap values, but a decayvalue of 1 signifies very low support for their BOPclade. With the recognition of the priority of thename Ehrhartoideae over Oryzoideae (see GPWG 2001),the acronym was changed from BOP to BEP and,therefore, the latter acronym will be used in this restof the paper. Aristida appeared sister to PACC plusBEP, a position that currently is not supported (Hiluet al. 1999; GPWG 2001). The Centothecoideae wasnested within the Panicoideae clade. Clark &Judziewicz (1996) re-instated subfamily status for theAnomochlooideae and redefined the Pharoideae as amonotypic subfamily.

Aiming at resolving phylogenetic relationships inthe Poaceae as a whole, Hilu et al. (1999) used the



rapidly evolving and information-rich matK genesequences from representatives of 60 grass genera toreconstruct a phylogeny for Poaceae. The study resolvedStreptochaeta and Anomochloa in a grade (divergingindividually) but not a clade sister to remaining Poaceae,and thus argued against the reestablishment (afterCaro 1982, see Table 1) of the subfamilyAnomochlooideae. This topology was later supported byMathews et al. (2000) and Zhang (2000) usingsequences from nuclear phytochrome B and theplastid rpl16 intron. Following these two clades, Pharusappeared sister to the rest of the Poaceae. Material forPuelia and Guaduella was not available and, thus, theirposition in grass phylogeny based on matK is yet to beestablished. The matK study did not recover the BEPclade since the Pooideae and Bambusoideae formed awell-supported clade but the Oryzoideae clade emergedsister to the PACC subfamilies with low support. Otherresults of the study are the segregation of the sistergroup relationship of a monophyletic Centothecoideae toa strongly supported Panicoideae, the sister grouprelationship of Aristida to remaining PACC taxa, a basalposition of a well-supported Brachyleytrum + Nardusclade to remaining Pooideae, a basal clade in theChloridoideae that included Uniola, Pappophorum andsome Eragrostis species, and the appearance oftraditional arundinoids in successive clades.

Molecular markers in the form of structural changesin genes/genomes can provide valuable phylogeneticinformation. Phylogenetically informative indels atthe 3 end of the matK gene in members of thePoaceae have been noted (Hilu & Alice 1999). Twoimportant indels are noteworthy. First is a six-basepair insertion uniting members of the PACCADclade. Second, a single base pair deletion at the endof the gene in Anomochloa and Streptochaeta that causeframe shift. This deletion is lacking in the remainingPoaceae but is shared with the sister familyJoinvilleaceae, thus confirming the basal-mostposition of these two genera.

The Hilu et al. (1999) publication was followed bythe Grass Phylogeny Working Group study aiming atreconstruction of phylogenetic relationships in thePoaceae, using information from six genes,chloroplast genome restriction sites andmorphological data (GPWG 2001). The mostparsimonious tree obtained from the combined dataanalysis resolved a monophyletic Streptochaeta andAnomochloa as first diverging grasses, followed by agrade of Pharus, Puelia plus Guaduella, and then twoclades representing the previously recognized BEPand PACCAD clades (see consensus tree in Fig. 11).Support for the above mentioned backbone brancheswas very strong (100% bootstrap) except for the BEP



Fig. 11. A consensus tree for the major grass lineages derived from various recent phylogenetic studies on the Poaceae. Incongruence intopologies of some clades, such as the PACCAD clade and the Anomochloa and Streptochaeta clades, is presented as polytomies.

clade that received 70% value. Analyses of individualdata sets revealed some incongruence in patterns ofphylogenetic relationships such as the emergence ofStreptochaeta and Anomochloa as a grade instead of aclade with rbcL data, and the divergence of theEhrhartoideae (Oryzoideae) after Pharus as sister toremaining grasses. Based on the emerging clades, thestudy recognized ten subfamilies: Anomochlooideae,Pharoideae, Puelioideae, Bambusoideae, Ehrhartoideae,Pooideae, Aristidoideae, Arundinoideae, Danthonioideae,and Centothecoideae. GPWG (2001) underscoredproblems related to the monophyly of theAnomochlooideae, the placement of Streptogyna, thenature of early diverging Pooideae, the lack of strongsupport for the Centothecoideae, and the low robustness(both in terms of resolution and support) in thestructure of the PACCAD clade.

Although the chloroplast genome was the focus inmolecular studies on the Poaceae, the ITS region(Hsiao et al. 1999) and the phytochrome gene family(Mathews & Sharrock 1996; Mathews et al. 2000) wereused in reconstructing phylogenies for the Poaceae.The Hsiao et al. (1999) parsimony tree resolved agrade of Streptochaeta, Pharus, Oryzoideae, andBambusoideae (paraphyletic BEP), followed by a splitinto Pooideae and PACCAD. Support for theserelationships was very strong (97 100% bootstrapvalues) except for the sister group relationshipbetween Bambusoideae and the remaining grasses. TheArundinoideae was monophyletic. The inferredtopology is rather good for a region that is difficult toalign at that level. The phytochrome B-basedphylogeny (Mathews & Sharrock 1996; Mathews et al.2000) was similar to that inferred from ndhF but withstronger support for the BEP clade.

Therefore, contributions from the AnalyticPhylogenetic period are gradually and cumulativelyproviding the building blocks for extensivephylogenetic hypotheses for the Poaceae. Althoughrobustness is still lacking in parts of the tree, theachievements are significant and the work is still inprogress. I have presented a consensus phylogeny formajor lineages of the Poaceae synthesized from variousphylogenetic studies, where incongruence ispresented as polytomies (Fig. 11).

Problems and prospects in grass systematicsIt is quite evident that a considerable progress ingrass systematics has been made. It is also evidentthat the current picture we have of the Poaceae is theresult of an accumulation of information andperspectives based on intuitive but acute insights andphenetic and phylogenetic analytical approaches. Itis rather remarkable to see observations made over ahundred years now substantiated by molecular workand phylogenetic reconstructions. The progress

made in grass systematics highlights the immenseefforts and contributions made by earlier grasssystematists, the concerted effort to synthesize theinformation that followed, and the commitment ofresearchers to incorporating recent advances inscience and engineering into resolving outstandingissues in grass systematics.

It is clear that overall gross morphology has led tothe lumping of otherwise distinct natural groupssuch as the case in the Pooideae of the two-subfamilysystem. However, as detailed information frommicrostructure and chromosome cytology becameavailable, the problem was recognized, andsegregation of new hierarchical entities in thePoaceae was proposed, leading to the initial multiplesubfamily systems of Avdulov (1931), Pilger (1954),Beetle (1955), Stebbins & Crampton (1961),Jacques-Felix (1962), Tateoka (1957), and others.The various sources of information fit well togetherto provide supporting evidence for newclassifications that prompted one of the first excitingperiods of progress in grass systematics. However, thenumber of subfamilies recognized varied (Fig. 1),stemming from subjective assessments of taxonomiclevels and group boundaries. Therefore, it is notsurprising to see the Phenetic Analytical periodshowing considerable stability in the number(around 5) and kind of subfamilies beingrecognized. Using varying numbers of diverse typesof characters has generated well-defined groups thatcan for the most part be easily discerned. Theperception of those researchers in the placement ofthe phenon lines skewed hierarchical definition oftaxa, leading to the splitting or lumping of clustersand, consequently, taxonomic units. An example ofthis is moving down the phenon line slightly in Hilu& Wright (1982) to circumscribe the cluster ofLygeum, Nardus and Diarrhena as a Nardoideae insteadof keeping them as part of the Pooideae, where thatposition corresponds to present-day definition ofthese taxa as basal Pooideae (Dring et al. this issue).Character scoring, as in the case of Watson et al.(1985), might have led to the partially distortedrelationships among grass subfamilies. Nevertheless,the major definable clusters of subfamilies remainedunaffected for the most part in the majority of thecontributions. The approach, however, lacked thephylogenetic component.

The introduction of cladistics and the recentadvances in molecular systematics have considerablyadvanced our understanding of grass systematics. Theseveral studies conducted at the whole family leveland within various subfamilies have provided newinformation on grass phylogenetics and ultimatelyserved as a foundation for defining its taxonomy atvarious levels. These studies represent only thebeginning for more detailed investigations needed in



the various subfamilies. Patterns of phylogeneticdivergence at the base of the family are welldocumented. However, lacking are robust phylogeniesamong members of the PACCAD clade, by far thelargest in the Poaceae, as well as relationships amongPooideae, Bambusoideae, and Ehrhartoideae and theiraffinities to the PACCAD clade.

The significance of recent advancements inunderstanding grass phylogeny is not restricted toincreased knowledge of taxon relationships but alsoin its application to furthering our knowledge oncharacter evolution. Evolutionary trends andpotential homoplasies have been deduced fromcurrent phylogentic hypotheses for various traits,such as morphological, anatomical, physiological,and cytogenetic characters (Soreng & Davis 1998;GPWG 2001; Hilu 2004, 2006). The high degree ofcongruence among phylogenies based on differenttypes of characters provides the opportunity forderiving a reliable consensus phylogeny that can beused to build evolutionary hypotheses for a numberof biological features in the Poaceae. The pattern ofchromosome evolution in the family has beeninconsistently depicted by various authors due to theimmense variability in basic chromosome numbers,the prominence of polyploidy in the family, and thesubsequent lack of a well-substantiated phylogeny.Current knowledge of grass phylogeny has providedan opportunity to map this character on a consensustree and to test previous hypotheses (Hilu 2004).Similarly, patterns of species diversity have also beenaddressed in a phylogenetic context using aconsensus grass tree to address species richness in ahistoric context and predict its potential causes (Hilu2006). Further studies in this area are encouraged asphylogenetic trees are established to increase ourunderstanding of the biology of organisms by placingit in evolutionary perspectives. As TheodosiusDobzhansky elegantly phrased the title of one of hisarticles: Nothing in biology makes sense except in the lightof evolution (Dobzhansky 1973).

The excitement of the application of molecularinformation in grass phylogenetics has resulted insome premature decisions about the definitions ofsubfamilies. Monophyly is a basic principle fordefining a natural group in systematic biology andphylogenetic methods provide healthy grounds forthe application of this dogma. However, using thiscriterion without critical assessments of the methodsand characters used in reconstruction, andrecognizing clades despite evidence to the contraryfrom other sources, can result in premature orerroneous taxonomic judgments. Low support forclades is a precautionary sign of their reality. Long-branch attraction can result in misleading affinitiesand taxon sampling and type of characters used canalso result in erroneous phylogenies as we have

experienced in angiosperm systematics (see Soltis etal. 2004). An example here is the re-establishmentof the Anomochlooideae (Clark & Judziewicz 1996;GPWG 2001) despite the lack of non-molecularsynapomorphies. Application of rigorous, model-based phylogenetic methods and the use of taxon-dense and character-diverse data sets are needed inthis case and other studies in the Poaceae.

The notable proliferation in the number ofsubfamilies during phylogenetic period also stemsfrom the taxonomic recognition of clades consistingof one or two genera at a subfamily level. Therecognition of a Pharoideae to encompass Pharus is anexample of this trend. Arguments can be made forgenera like Micraira that appears sister to thePACCAD clade (GPWG 2001) and for some distinctlineages sister to major grass subfamilies, as in thecase of Eriachne. Confining the subfamily definition tolineages with larger groups of genera makes thetaxonomy of this large family more comprehensible.These monotypic small tribes are isolated islands, andislands ought not to be treated as continents. Movingin the direction of subfamily proliferation is a trendthat has the danger of acceleration with potentialnegative consequences on grass systematics.

Prospects in grass systematicsThree important factors provide even brighterprospects for advancement in grass systematics. One,Morpho-Data banks are not only available on the webfor the Poaceae, but are becoming increasingly user-friendly. Notable among these are the Grass Genera ofthe World by Watson & Dallwitz (1992 onwards:http://delta-intkey.com/grass/) and Kew World GrassSpecies by W. D. Clayton, K. T. Harman & H.Williamson (http://www.rbgkew.org.uk/data/grasses-db/ident.htm). The second factor is the accumulationof an immense amount of information from genesequences in GenBanks that provides a wealth ofcharacters for exhaustive analysis at all taxonomiclevels in the Poaceae. Sequencing whole genomes isanother major source of molecular information thatwill impact our knowledge on grasses. The availabilityof these data is paving the way for the third factor,namely the exhaustive analyses of individual andcombined data sets of molecular and non-molecularcharacters. The continued improvement in computerpower and the availability of Super Computersmakes the task of rigorous analyses of large data lessproblematic and more practical in terms of time andfunding. The new trend in establishing super trees isalso promising where tree matrices are used in placeof data matrices, and thus overcoming problemsrelated to incongruence in taxon representationsencountered in different analyses and its consequentproblem of missing data.



To enhance the progress in grass systematics, agenomic DNA bank is most needed. A number ofgenera have narrow geographic distribution orsimply are endemics, some species are not easilyaccessible, and material is difficult to find forcertain taxa that appear to occupy crucial positionsin the phylogeny of grasses in some analyses.Unavailability of plant or DNA samples is anobstacle to further studies at the molecular levelthat are using different genomic regions or forthose attempting to expand sample size. Usinggenes from different genomic regions as well asdense sampling are important criteria for achievingreliable phylogenies (see Soltis et al. 2004).Establishing a Poaceae genomic DNA bank that canprovide DNA samples with no restrictions butpossibly with minimal fees to recover some of thecosts would strongly enhance the progress in grassmolecular systematics in particular, and grasssystematics in general.

The next two or three decades are expected tobring a wealth of insight into grass systematics. It is anexciting period for us to experience and contributeto as we observe the ongoing molecular andinformatics revolution impacting the biologicalsciences in general and systematic biology inparticular. However, information from morphology,anatomy and physiology remain central to ourunderstanding of grass systematics, and currentbioinformatics methods should enhance our utility ofboth molecular and non-molecular information.

AcknowledgmentsThe author thanks two anonymous reviewers for theirvaluable comments on the manuscript.

ReferencesAdanson, D. (1763). Familles des Plantes. Vol. 2.

Paris, Vincent.Avdulov, N. P. (1931). Karyosystematic Studies in the

Grass Family. Bull. Applied Bot. Genet. Pl. Breed.,Supplement 44: 1 352.

Barker, N. P., Linder, H. P. & Harley, E. H. (1995).Polyphyly of the Arundinoideae (Poaceae): evidencefrom rbcL sequence data. Syst. Bot. 20: 423 435.

, , (1999). Sequences of the grass-specific insert in the chloroplast rpoC2 geneelucidate generic relationships of theArundinoideae (Poaceae). Syst. Bot. 23: 327 350.

Baum, B. R. (1987). Numerical taxonomic analyses ofthe Poaceae. In: T. R. Soderstrom, K. W. Hilu, C. S.Campbell, M. E. Barkworth (eds), Grass systematicsand evolution, pp. 334 342. SmithsonianInstitution Press, Washington, DC. USA.

Beetle, A. A. (1955). The four subfamilies of theGramineae. Bull. Torrey Bot. Club 82: 196 197.

Bentham, G. (1881). Notes on Gramineae. J. Linn.Soc., Bot. 19: 14 134.

(1887). Flora AustraliensesBrown, R. (1814). General remarks, geographical and

systematical, on the botany of the Terra Australis,Appendix 3. Pp. 580 583. In: M. Flinder (ed), AVoyage to Terra Australis; Undertaken for thePurpose of Completing the Discovery of that VastCountry, and Prosecuted in the Years 1801, 1802,1803. W. Blumer, London.

(1878). Flora Australiensis: A Description of thePlants of the Australian Territory, vol. 3. L. Reeves& Co., London.

Caro, J.A. (1982). Sinopsis taxonmica de lasgramneas argentinas. Dominguezia 4: 1 51.

Celakovsk, L. (1889). ber den rchenbau derBrasilianischen Grasgatung Streptochaeta Schrader.Sitzungsberichte der Kniglichen BhmischenGesellschaftder Wissenschaften. Mathematisch-naturwissenschaftliche Classe 3: 14 42, Table 2.[Unpublished translation by Mrs. Geti Saad,1974, for the Smithsonian Institution and theNational Science Foundation, Washington, D. C.;in the files of the McClure Bamboo Library,Department of Botany, Smithsonian Institution,Washington, DC.

Clark, L. G., Zhang, W. & Wendel, J. F. (1995). Aphylogeny of the grass family (Poaceae) based onndhF sequences. Syst. Bot. 20: 436 460.

& E. J. Judziewicz. (1996). The grass subfamiliesAnomochlooideae and Pharoideae (Poaceae).Taxon 45: 641 645.

Clayton, W. D. (1978). Gramineae. In: V. H. Heywood(ed), Flowering Plants of the World. MayflowerBooks, New York, USA.

& Renvoize, S. A. (1986). Genera Graminum.Grasses of the World. (Kew Bull., Addit. Ser. XIII),389 pp. HMSO, London, UK.

Clifford, H. T. (1965). The classification of thePoaceae: a statistical study. Pap. Dept. Bot. Univ.Qld. 4: 243 253.

& Goodall, D. W. (1967). A numericalcontribution to the classification of the Poaceae.Aust. J. Bot. 15: 499 519.

, Williams, W. T., & Lance, G. N. (1969). Afurther numerical contribution to the classificationof the Poaceae. Aust. J. Bot. 17: 119 131.

Davis, J. I. & Soreng, R. J. (1993). Phylogeneticstructure in the grass family (Poaceae) as inferredfrom chloroplast DNA restriction site variation.Am. J. Bot. 80: 1444 1454.

Dobzhansky, Th. (1973). Nothing in biology makessense except in the light of evolution. Amer. Biol.Teacher 35: 125 129.



Doebley, J., Durbin, M., Golenberg, E. M., Clegg M.,& Ma, D. P. (1990). Evolutionary analysis of thelarge subunit of carboxylase (rbcL) nucleotidesequence among the grasses (Gramineae).Evolution 44: 1097 1108.

Dring, E., Albrecht, J. T, Hilu, K. W., & Rser, M.Phylogenetic relationships in the Aveneae/Poeaecomplex (Pooideae, Poaceae). Kew Bull. (this issue).

FAOSTAT. (1999). Database on line,http//:apps.fao.org/

GPWG, Grass Phylogeny Working Group (2001).Phylogeny and subfamilial classification of thegrasses (Poaceae). Ann. Missouri Bot. Gard. 88: 373 457.

Hamby, R. K. & Zimmer, E. A. (1988). RibosomalRNA sequences for inferring phylogeny within thegrass family (Poaceae). Pl. Syst. Evol. 160 : 29 37.

Harz, C. D. (1885). LandwirthschaftlicheSamenkunde. Verlag von Paul Parey, Berlin.

Hilu, K. W. (2000). Contributions of Prolamin SizeDiversity and Structure to the Sytematics of thePoaceae. In: S. W. L. Jacobs & J. Everett (eds),Grasses: Systematics and Evolution, Vol. 2, pp. 241 247. CSIRO, Melbourne.

(2004). Phylogenetics and chromosomalevolution in the Poaceae (grasses). Austr. J. Bot. 52:13 22.

(2006). Skewed distribution of species numberin grass genera: is it a taxonomic artifact? In: T.Hodkinson, J. Parnell & S. Waldren (eds), Towardsthe Tree of Life: the taxonomy and systematics oflarge and species rich groups. CRC Press,Washington D.C.

& Wright, K. (1982). Systematics of Poaceae: ACluster Analysis Study. Taxon 31: 9 36.

& Esen, A. (1988). Prolamin size diversity in thePoaceae. Biochem. Syst. & Ecol. 16: 457 465.

& L. Sharova (1998). Characterization of 10kDaProlamin Genes in Phyllostachys aurea Riv.(Bambusoideae, Poaceae). Am. J. Bot. 85: 1033 1037.

& Alice, L. A. (1999). Evolutionary implicationsof matK indels in Poaceae. Am. J. Bot. 86: 1735 1741.

, & Alice, L. A. (2001). A phylogeny ofChloridoideae (Poaceae) based on matK Sequences.Syst. Bot. 26: 386 405.

, , & Liang, H. (1999). Phylogeny of Poaceaeinferred from matK sequences. Ann. Missouri Bot.Gard. 86: 835 851.

Hitchcock, A. & Chase, A. (1950). Manual of theGrasses of the United States. 2nd. Ed. U.S.D.A.Misc. Publ. 200, USA.

Hsiao, C., Jacobs, S. W. L., Chatterton, N. J., & Asay, K.H. (1999). A molecular phylogeny of the grassfamily (Poaceae) based on the sequences of nuclearribosomal DNA (ITS). Aust. Syst. Bot. 11: 667 688.

Jacques-Flix, H. (1962). Les Gramines dAfriqueTropicale. Bull. Sci. Inst. Rech. Agron. Trop. 8: 1 345.

Kellogg, E. A. & C. S. Campbell (1987). Phylogeneticanalysis of the Gramineae. In: T. R. Soderstrom, K.W. Hilu, C. S. Campbell, & M. A. Barkworth (eds),Grass systematics and evolution, pp. 310 322.Smithsonian. Institution Press, Washington, DC.

Mathews, S. & Sharrock, R. A. (1996). Thephytochrome gene family in grasses (Poaceae): aphylogeny and evidence that grasses have a subsetof the loci found in dicot angiosperms. Mol. Biol.Evol. 13: 1141 1150.

Mathews, S., Tsai, R. C. & Kellogg, E. A. (2000).Phylogenetic structure in the grass family(Poaceae): Evidence from the nuclear genephytochrome B. Am. J. Bot. 87: 96 107.

Paroid, L. R. (1961). La taxanoma de las GramineaeArgentinas a la luz de las investigaciones mssrecintes. Adv. Bot. 1: 125 129.

Pilger, R. (1954). Das System der Gramineae. Bot.Jahrb. 76: 281 384.

Roshevitz, R. Yu. (1937). Zlaki: Vvedenie v Izuchenie iKormovykh Zlakov. Selkozgiz, GosudarstvennoeIzdatelstov Kolkhoznoi I Sovkhoznoi Literatury,Moskva, Russia. Grasses: An Introduction to theStudy of Fodder and Cereal Grasses. Translatedand published for the Smithsonian Institution in1980 by the Indian National DocumentationCenter. New Delhi, India.

Schuster, J. 1910. ber die morphologie dergrasblte. Flora 100: 213 266.

Shantz, H. L. (1954). The place of grasslands on theearths cover of vegetation. Ecology 35: 143 145.

Soderstrom, T. R., Hilu, K. W., Campbell, & C. S.,Barkworth, M. E. (eds). (1987). Grass systematicsand evolution. Smithsonian Institution Press,Washington, DC., USA.

Soltis, D. E., Albert, V. A., Savolainen, V., Hilu, K.,Qiu, Y-L., Chase, M. W., Farris, J. S., Stefanovi, S.,Rice, D. W., Palmer, J. D., &. Soltis, P. S. (2004).Genome-scale data, angiosperm relationships, andending incongruence: a cautionary tale inphylogenetics. Trends Plant Sci. 9: 477 483.

Soreng, R. J., & J. I. Davis. (1998). Phylogenetics andcharacter evolution in the grass family (Poaceae):Simultaneous analysis of morphological andchloroplast DNA restriction site character sets. Bot.Rev. 64 1 85.

Stebbins, G. L. (1956). Cytogenetics and evolution ofthe grass family. Am. J. Bot. 43: 890 905.

Stebbins, G. L. (1982). Major Trends of Evolution inthe Poaceae and Their Possible Significance. In:J. R. Estes, R. J. Tyrl, & J. N. Brunken (eds),Grasses and Grasslands: Systematics and Ecology,pp. 3 36. University of Oklahoma Press,Norman.



, & Crampton, B. (1961). A suggested revision ofthe grass genera of temperate North America. In:Recent Advances in Botany, Vol. 1, pp. 133 145.Univ. Toronto Press, Toronto, Canada.

Tateoka, T. (1957) Miscellaneous papers on thephylogeny of Poaceae, X: Proposition of a newphylogenetic system of Poaceae. J. Jap. Bot. 32: 275 287.

Tsvelev, N. N. (1976). Grasses of the Soviet Union.Part I. Translated in 1983 by B. R. Sharma.Amerind Publ. Co. Pvt. Ltd., New Delhi.

(1989). The system of grasses (Poaceae) andtheir evolution. Bot. Rev. 55: 141 204.

Watson, L. & Dallwitz, M. J. (1992). The Grass Generaof the World. Wallingford, Oxon, UK: CABInternational.

, Clifford, H. T., & Dallwitz, M. J. (1985). Theclassification of the Poaceae: subfamilies andsupertribes. Aust. J. Bot. 33: 433 484.

Zhang, W. (2000). Phylogeny of the grass family(Poaceae) from rpl16 intron sequence data. Molec.Phylogenet. Evol. 15: 135 146.



/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 300 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 150 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 600 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False

/Description > /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ > /FormElements false /GenerateStructure true /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles true /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /NA /PreserveEditing true /UntaggedCMYKHandling /LeaveUntagged /UntaggedRGBHandling /LeaveUntagged /UseDocumentBleed false >> ]>> setdistillerparams> setpagedevice

a century of progress in grass systematics

Documents