Ideas on the classification of grasses are in a constant state of flux. Current work in this field is being conducted in many centres as can be seen from the proceedings the symposia on the classification and evolution of grasses from 1986 onwards (Soderstrom, Hilu, Campbell and Barkworth 1987; Jacobs and Everett 2000; J.T.Columbus et al 2007; Seberg et al. 2010).

 The subfamilies recognised are a reflection of the prevailing consensus of the Grass Phylogeny Working Group recently published in New Phytologist (GPWG II 2011). This classification builds on the first classification of the GPWG published in the Annals of the Missouri Botanical Garden (GPWG 2001).

GWPG II  (2011)   New Phytologist 193: 304-312 (2011)  

GPWG (2001)    Compiled from Annals of the Missouri Botaical Garden 88(3): 395-397, Fig 3 (2001)


There are currently, and have always been, differences of opinion regarding generic circumscription in the Poaceae, or indeed in any plant family. For the last 20 years the main generic reference for grasses has been Genera Graminum of Clayton and Renvoize (1986), a subjective compilation of 651 genera from the accumulated knowledge of the collections, library, curation and expertise at the grass herbarium of the Royal Botanic Gardens Kew. A few years after Genera Graminum another book on world grass genera was published (Watson & Dallwitz, 1992), in which detailed descriptions of 785 genera, using 496 characters were produced by computer software of the DELTA system (Dallwitz, 1980; Dallwitz, Paine & Zurcher, 1993 onwards), from data that had been gathered for twenty years by the senior author.  The data for this database is currently maintained on a website (Watson & Dallwitz, 1992 onwards  


Prior to these two works the attempts to enumerate the world grass genera reflected the times in which they were compiled, both in terms of the classifications that existed and the regions of the world that had been explored and collected. A summary of all global accounts of grass genera is presented in Table 1.

Table 1.  Accounts of World Grass Genera from Linnaeus until present day

Number of Genera

Number recognised in GrassWorld

% Representation in GrassWorld

% of  Total



      1.Linnaeus 1753





      2.Palisot de Beauvois 1812





      3.Trinius 1820





      4.Kunth 1829





      5. Steudel 1855





      6. Bentham 1882





     7. Bentham & Hooker 1883





      8. Hackel  1887





     9.  Bews 1929





     10. Roshevitz 1937





         11. Pilger 1954 (no bamboos)





        12.Tzvelev 1989





    13. Clayton & Renvoize 1986





    14.  Dallwitz & Watson 1988





       15. GrassBase(Clayton & Harman 2006)





       16. GrassWorld (Simon &  Healy 2007)








The earliest world compilation of grass genera, understandable in today’s nomenclature, is the entry for the family in the Species Plantarum of Linneaus (1753), where 39 grass genera were described and are all still recognised today. Following Linnaeus the next botanist to propose a classification of the grasses that became widely accepted was that of Robert Brown in his Prodromus (Brown, 1810, 1814; Clark, 2004), but this work was restricted in geographical scope to Australia. Further developments in grass classification in the 19th century are discussed in the introduction to the classic essay on grasses by Bentham (1882), in which he presented his own overview of grass genera on a global scale and reviewed those that preceded him. These were the accounts of Palisot de Beauvois (1812), Trinius (1820), Kunth (1829) and Steudel (1855). The numbers of genera in these historical publications vary from 80 (Kunth, 1829) to 570 (Steudel, 1855) (Table 1), with Bentham himself accounting for 275 (Bentham, 1882), expanded a year later to 311 (Bentham & Hooker, 1883). The low number for Kunth was probably due to the fact that his account was prepared too hastily (Bentham, 1882) while the high number of Steudel was due to the fact that he was an extreme splitter and in the opinion of Bentham very careless in his work, giving rise to the comment by him that Steudel’s work was “a perfect chaos”. Soon after Bentham’s system was published, Hackel (1890 published a very similar account of grass genera in Engler and Prantl’s Pflanzenfamilien, later translated into English (Hackel, 1896). Although the total number of grass genera is similar to that of Bentham and Hooker, a larger number of Hackel’s genera are still recognised today, compared to the number for Bentham and Hooker (Table 1).

In the early twentieth century the two main global accounts of grass genera were those of Bews (1929) with 481 genera and Roshevits (1937) with 563 genera, the latter Russian work not well known for a long time until translated into English (Roshevits, 1980). The increase in the number of new genera in these two accounts from that of Hackel was 35% for Bews and 44% for Roshevits. By the 1950’s the global generic number for the grasses was still only 555 (Pilger 1954), but this number did not include the bamboos.

The two contemporary accounts of the grass genera on a global scale (Clayton & Renvoize, 1986 and Watson & Dallwitz, 1992 onwards) were published within ten years of each other, and yet have a difference of 134 genera. The rationale for this difference is well explained in the introduction to Watson and Dallwitz’s book, where the authors state that due to conflicting views of taxonomists on generic limits, they made a decision to accept the narrower circumscriptions of genera, where they existed, for the reason that is was easier to “coalesce existing descriptions than to subdivide them.” A third classification proposed about the same time as the two systems above was the classification of Tzevelev (1989). In it he listed 905 genera arranged within 41 tribes, placed in two subfamilies Bambusoideae (14 tribes) and Pooideae (27 tribes). The reason that there are considerably more genera than those of Clayton & Renvoize or Watson & Dallwitz, can probably be attributed to a tendancy of extreme spitting by Russian botanists. For example there are 191 genera listed by Tzvelev that are placed in synonymy with other genera in contemporary systems (GrassWorld), and there are four genera (Fingerhuthia, Cyathopus, Dichelachne and Dissochondrus) that are listed twice under differernt numbers and in different tribes.

There are currently two compilations of the world’s grasses at species level being assembled. One, GrassBase (Clayton, Vorontsova, Harman & Williamson 2012 and ongoing at, is essentially a set of online morphological descriptions for the grass species of the world compiled at the grass herbarium of the Royal Botanic Gardens Kew since 1997, using DELTA format. The most recent addition to this set is the list of online descriptions of grass genera compiled from the constituent species by a program gesumm written by Mike Dallwitz (Dallwitz, 2007 onwards). The second global database is GrassWorld (Simon, 2007a, Simon & Healy, 2007), from which this Scratchpad is produced,  is being compiled at the Queensland Herbarium and is essentially the morphological information of GrassBase together with 14 ancillary non-morphological characters. The ultimate aim of GrassWorld is an interactive online information system and key of world grasses. The information is currently in four languages – English, German, Spanish and French. The interactive  key will initially be released in the Intkey module of DELTA, but the possibility exists for it to be released in the Lucid format (Centre for Pest Information Technology and Transfer, 1999), similar to AusGrass (Sharp & Simon, 2002). The latter is essentially an enhanced subset of an early version of GrassBase.

Table 2.  Numbers of Genera and Species in GrassWorld   (2012)




Basal Subfamilies

(Anomochlooideae, Pharoideae, Puelioideae)

































Presently GrassWorld comprises 823 genera and 12,096 species compared to 702 genera and 11,086 species in a 2008 release of GrassBase. The greater number of genera in GrassWorld from those in Genera Graminum can be considered under two categories - those described as new since 1986 and those placed in synonymy in Genera Graminum but since resurrected. Some of these genera have been accepted in GrassBase and others not. Likewise many of the genera in Watson and Dallwitz (1992) are not accepted in GrassWorld. There are also a few genera from Genera Graminum not recognised in GrassWorld.


The selection and use of characters. – Of prime importance as to how we define the taxa we present in our classifications, be they species, genera or higher ranks, is how we circumscribe them and key them out in terms of their characters. An excellent overview of the sorts of taxonomic characters that exist (Davis & Heywood, 1963) indicates that the prime source of characters available in plant taxonomy come from morphological and anatomical features, although other types of characters discussed include cytological and chemical. In the grasses cryptic characters became to be used from the mid 20th Century, when it became apparent that reliance only on morphological floral characters was not producing satisfactory classifications, particularly at the upper levels. Characters of particular value in this regard were characters from cytology (Avdulov, 1931), leaf anatomy (Prat, 1936, 1960; Brown, 1958) and embryology (Reeder, 1957). Up to this time only phenetic characters were taken into consideration in making classifications, even when talking about “natural” classifications (Davis & Heywood 1963). After 1966 the impact of Hennigian phylogenetic theory (Hennig 1966) resulted in a paradigm shift in taxonomic practice, with the ability to analyse shared derived character states; at first morphological characters were used and subsequently molecular sequences from both the chloroplast and nuclear regions. Following from this work, different categories of character states had to be taken into consideration when making phylogenetic classifications – whether they were synapomorphic (unique), symplesiomorphic (shared primitive) or homoplasious (those arising more than once). There appear to be two major areas of contention among contemporary taxonomists regarding the application of phylogenetic tools to classification - the search for and the recognition of monophyletic groups; furthermore, if these groups are produced by molecular methods should they be also be able to be recognisable by observable morphological features?

 Monophyly and paraphyly. -- In the last decade there has been much discussion in taxonomic and phylogenetic literature on whether paraphyletic taxa should (Brummitt, 1997, 2002, 2003, 2006; Hörandl 2007; Nordal & Stedje 2005; Sosef 1997, van Wyk 2007) or should not (Dias, Assis & Udulutsch 2005; Ebach, Williams & Morrone 2006; Freudenstein 1998; Nelson, Murphy & Ladiges 2003; Pfeil & Crisp 2005; Potter & Freudenstein 2005) be given recognition. In addition there seems to be a difference of opinion as to how a taxonomist should present their classifications (Diggs & Lipscomb 2002), with two opposing schools being evident – one is evolutionary taxonomy, which allows for paraphyly and the other is phylogenetic systematics, following strict monophyly. Evolutionary taxonomy is character dependent, whereas phylogenetic systematics may or may not have corroborative morphology to support clades derived from molecular data (Simon 2007b).

Parsimony.-- Although the concept of monophyly appears theoretically logical, following the principles of Hennig (1966), one of its basic flaws appears to be the interpretation of the branching patterns of a cladogram derived by parsimony. Conventionally this has been presented as being the only factor to be considered in understanding a phylogeny (Stuessy & Konig 2009). However, as pointed out recently by Hörandl (2006), “a strict application of monophyly for grouping of taxa is problematic, because the commonly used tree-building methods result in a too strong abstraction and a too simplified visualization of evolutionary processes”. Central to this model is the interpretation of the dichotomous branching pattern of a cladogram with regards to ancestor-descendant relationships (Simon 2000). Evolutionary processes operate without the extinction of parental groups when new taxa originate, whereas the Hennigian model does not allow for this. This is well illustrated by Sosef in his discussion of two hierarchical models – monophyletic and Linnean – in his figs 2a and 2b (Sosef, 1997); neither model is wrong, both depicting how two types of classification can be applied to the same taxonomic group; one accepts paraphyly and monophyly and the other only monophyly. Many contemporary classifications are being presented as a compromise between the two. Furthermore dichotomy by itself does not readily account for anagenesis, reticulate evolution, hybridisation and polyploidy. Cladistic programs are unable to readily detect reticulate patterns of relationship (Sosef 1997), a situation probably more common in nature than the models presented by the dichotomous branches on a cladogram. In the grasses, with which this paper is mainly concerned, at least 80% of grass species are of polyploid or aneuploid origin (de Wet 1987). How this process is depicted within a cladistic classification, particularly in cases where well known compilospecies such as Bothriochloa bladhii (Retz.) S.T.Blake hybridizes  naturally with other species of Bothriochla and with species of the related genera Capillipedium and Dichanthium (de Wet & Harlan 1966), is problematical.

Another critique of parsimony being the sole explanation for the evolutionary process and being used as the framework for classification is discussed by Michael Lee (Lee, 1999, 2002, 2003, 2004). Some of the relevant statements of Lee’s in connection with these issues are

“Inclusion of phylogeny-dependent principles in systematic studies is circular, since such principles have no external empirical support but are themselves derived from systematic studies.” (Lee 1999);

“When ordering biological diversity, hierarchical ordering procedures (such as cladistics) can only be preferred over other systems if one accepts not only that the entities change (evolve), but that a particular pattern of evolutionary change prevails (branching and divergence). In choosing to use cladistics, therefore, one makes very specific evolutionary assumptions.” (Lee 2002);

“The idea of monophyly, and recognition of clades, can only readily be applied to nonreticulating entities with hierarchical relationships. Reticulation has shown to be problematic around the species level and at higher levels, but more likely to be prevalent at lower levels.” (Lee 2003).

 The fourth paper (Lee 2004) is more concerned with issues to do with phylogeny and classification based on molecular data and lists a number of useful papers on this subject by authors who either advocate a new taxonomic system based solely on DNA (Tautz et al. 2003), while others regard this as a backward step (Seberg et al. 2003; Lipscomb et al. 2003; Mallett & Willmott 2003).

Taxonomy today. -- I am of the opinion that a practical solution to aim for is a balance between evolutionary taxonomy and phylogenetic systematics, and that taxonomists present a uniform methodology to the users of our classifications in the way they operate.  As stated recently “taxonomy should be a science that earns its respect from users as well as fellow scientists” (Farjon, 2007) and it is therefore imperative that systematists are able to present a unified front all parties. A possible compromise between evolutionary taxonomy and phylogenetic systematics has been termed phyletics (Stuessy, 1997, 2009; Stuessy & König 2008), but computer software that is able to satisfactorily blend the elements of cladistics, patristics and phenetics is as yet undeveloped (Hörandl 2006).

Monotypic genera. -- When a taxonomist is presented with material of what appears to be a new species that cannot be readily assigned to any known genus, a practical difficulty arises. If this taxon is considered to be in a rare or endangered category it is important to conservationists, legislators and end-users that it be formally described as soon as is practically convenient, for reasons of communication. This is what transpired with the two new grass genera Alexfloydia and Dallwatsonia I described from Australia (Simon 1992). At the time these two genera were described as new, after attempts to place them with existing genera on the basis of phenetic morphological chraracters were not successful. Two other supposed new genera (Cliffordiochloa and Fasciculochloa) were erected using the same methodology (Simon 1992; Simon & Weiller 1995) but they were subsequently found to be introduced species of the genus Steinchisma from the New World (Simon 1999; Simon  2003). The reasons for these discrepancies was that the coding of characters in two data sources, under alternative names for the same taxon, disagreed for several characters. This illustrates the need for a standardisation of morphological characters so that they can be interpreted uniformly.

A meeting held in Melbourne in 2003, to assist in the preparation of a Census of Australian plants, recommends as a first guideline (Entwisle & Weston 2005) that, where possible, named taxa should be monophyletic based on current reliable evidence. However the guidelines do give recognition to paraphyletic and even polyphyletic genera as an interim step, pending further research on the group in question, in order to minimise taxonomic change across Australia (Guideline 2). However, there is a strong possibility that names of grass genera are unlikely to have a stable agreed nomenclature for the foreseeable future due to the increasing number of phylogenetic studies and possible changed classifications in the grass family. Another view considered at the 2003 meeting, but rejected for the purpose of the preparation of an Australian census, was that monotypic taxa are best avoided. They are regarded by pure cladists as of little scientific value (Platnick 1976; Freudenstein 1998) as they say that the concept of monophyly cannot be applied to a single terminal taxon, essentially what a monotypic genus is. The new grass genera I described belong to this category, as do the majority of grass genera (293 listed as monotypic in Clayton & Renvoize 1986). This illustrates a chasm between the theoretical desideratum of cladistics to avoid the recognition of monotypic genera and the fact that most present day grass genera (35% fide Hilu 2006) are monotypic. A similar situation prevails for the four families with more species than the grasses (Asteraceae, Leguminoseae, Orchidaceae and Rubiaceae) as well as 14 other large families (Clayton 1972).

An opposing view regarding the non-recognition of monotypic genera is that of Wiley (1977), who states that all monophyletic taxa must have been monotypic at the time of their origin. Although many grass genera have been described as monotypic, the situation often arises where a single species is described for a new genus, to be followed by the description of a second species some time later.  

 Generic concepts. -- The understanding and use of the generic rank has been examined from several aspects (Clayton 1972; Clayton 1983; Stevens 1985). Although the concept is traditionally considered a subjective grouping of morphologically similar species and has been used in folk taxonomies and classifications since pre-Linnean times, the biological interpretation of the genus is far from clear (Clayton 1983). Why have there been so many monotypic genera described for those species that have not been able to be grouped with other species? Are they all of a similar nature to the monotypic Australian genera that I established on phenetic principles, based on not being able to satisfactorily place them in existing genera? If the establishment of monotypic genera has traditionally been a taxonomic practice in former times, it is contrary to current cladistic views on monotypic taxa. If this viewpoint is to prevail, major upheavals are forecast to accommodate the subsuming of monotypic genera in the phylogenetic classifications of the future. Two recent quotes highlight this thinking:

“I suspect over the next couple of decades there will be a perhaps unprecedented period of lumping and splitting of genera, driven in large part by the wide availability of computing machines and the ease and low cost of obtaining molecular data” (Columbus, pers. comm.)

“Over the next ten years molecular research will cease to be a controversial issue in taxonomy and become a staple----. In the future molecular data will be so easy to gather that it will not require a human hand, much less a university trained researcher. Potentially every new collection will be examined with molecular tools before anyone sorts or determines it” (De Priest, 2000).

The fluctuation in generic size in plants in general have been examined in an excellent recent paper by Humphreys and Linder (2010), and what they say is very much applicable to the grasses. There seems to have been fluctuating trends of having small or large genera; the reasons of either the splitting of large genera to satellite genera or conversely the subsuming of small genera back into large genera to avoid paraphyly are discussed for many plant groups. This includes the concept of ‘good’ genera in terms of ‘stability’ and ‘predictiveness’, although it is not clear where exactly the authors stand in relation to paraphyly, with comments such as:

“efforts to define a genus that reflects evolutionary patterns should not override those that ensure its usefulness.” and “monophyly is important only in so far as it increases the chances of a genus being both stable and predictive”

contrasting with

“a classification that does not convey evolutionary units may be misleading since it is likely to be interpreted as doing so”

 Sometimes new species are discovered or sent to herbaria when there is a need to describe them for the purposes of conservation. Such a case arose recently for a grass collected from north Queensland. Initially this new species was unable to be placed in any known genus on the basis of it having very short lower glumes and very long upper glumes. I was considering erecting a new genus for this grass, using phenetic morphological characters as before. However on closer inspection, it was noticed that a few of the racemes terminated in bristle-like appendages, a character not previously observed, and also had smooth upper lemmas; these two characters necessitated placing it in the genus Pseudoraphis. Another example is the current work on a chloridoid grass from sandstone areas of inland Queensland; it is superficially similar to known species of the panicoid genus Neurachne, where it was originally tentatively placed, but following further morphological and molecular work (Macfarlane and Columbus, pers. comm.) it has been shown to have an affinity with the chloridoid genus Uniola, a genus that does not have any native representation in Australia. More species that cannot be placed in established genera will be discovered. I maintain that it is preferable that they be given names as monotypic genera as an interim measure, until such a time as phylogenetic work is done to establish where they belong. This fulfils the present need for them to have an effective label for communication. Another example of a genus that was published as new on the basis of its economic importance and before being classified is the genus Suddia Renvoize (Renvoize et al. 1984). It has been included in the Pharoideae (Clayton & Renvoize 1986; Watson & Dallwitz 1992), but it was not accepted as belonging to this subfamily when it was formerly described by Clark & Judziewicz (1996). These authors suggest that Suddia could be a member of the Erhartoideae on the basis of anatomical data, as normal spikelets have yet to be seen, due to their infection of the spikelet with a smut; as yet the genus still is to be phylogenetically placed on the basis of molecular data, a recent suggestion being that it belongs in the basal pharoid lineage (Kellogg pers. comm.).

Basal lineages. -- The early history of the use of molecular techniques to investigate grass phylogeny is well documented (summary in Simon, 2007b). Prior to this work (Stebbins 1956; Watson et al. 1985; Clayton & Renvoize 1986; Watson & Dallwitz 1992) the subfamily Bambusoideae was circumscribed to either include or it was placed near to some genera that are today conventionally called “basal lineages” of the Poaceae; they include three small subfamilies Anomochlooideae (Anomochloa Brongn. and Steptochaeta Schrad. ex Nees), Puelioideae (Guaduella Franch. and Puelia Franch.) and Pharoideae (Pharus P.Browne, Leptaspis R.Br. and Scrotochloa Judz.).

When this group of three subfamilies was first investigated, together with a wide sampling of the whole family (Clark et al 1995), it became clear that they were phylogenetically basal to all other grasses, even though no morphological synapomorphies were detected. This was followed by a formal recognition of the subfamily Anomochlooideae, with the tribes Anomochloeae and Streptochaeteae, and the subfamily Pharoideae (Clark & Judziewicz 1996); later, when the genera Puelia and Guaduella were added to the analysis and found to form a separate basal clade from the other two subfamilies, they were formerly described as subfamily Puelioideae (Clark et al. 2000).

Scratchpads developed and conceived by (alphabetical): Ed Baker, Katherine Bouton Alice Heaton Dimitris Koureas, Laurence Livermore, Dave Roberts, Simon Rycroft, Ben Scott, Vince Smith