Malvaceae Info: History of Classification (Draft)
Malvaceae Info (Home)
The table below gives a rough summary of how malvaceous taxa have been divided into families in the past, in terms of the elements Malvaceae s.s. (Malv.), Matisieae (Mat.), Bombacoideae s.s. (excluding Matisieae & Fremontodendreae) (Bomb.), Durioneae (Dur.), Fremontodendreae (Frem.), Helictereae (Helict.), Sterculioideae (Sterc,), Dombeyoideae (Domb.), Lasiopetaleae (Lasio.), Hermannieae (Herm.), Byttnerieae (Bytt.), Theobromateae (Theo.), Brownlowioideae (Brown.), Grew. (Grewioideae), Tilioideae (Til.), and also showing the inclusion of Muntingiaceae (Munt.), Elaeocarpaceae (Elaeo.), Prockieae (Prock.), Bixaceae (Bix.), Plagiipteraceae (Plag.) and Sphaerosepalaceae (Sphaer.) within otherwise malvaceaous families. This is an oversimplication, as not only have these elements been shuffled between families, but genera have been shuffled around. For more details see the descriptions of classificaitions following the table.
The families most commonly recognised were Malvaceae, Bombacaceae, Sterculiaceaes, Byttneriaceae and Tiliaceae. These are shown by red, orange, green, cyan, and blue cells respectively. Yellow represents Hermanniaceae (or Melochiaceae) and magenta cells, in various shades, other segregant families, grey cells non-malvaceous families such as Tithymalaceae, Muntingiaceae, Elaeocarpaceae, Flacourtiaceae, Bixaceae, Plagiopteraceae and Sphaerosepalceae, white cells groups which were not known at the time, and blank cells taxa for which I lack data, either because they were not known when the classification was made, were not in the scope of the work in question (e.g. a regional flora), or because I haven't found the relevant bit of the work.
Other segregate families include Berryaceae Doweld (2001), Cacaoaceae Augier ex T. Post & Kuntze (1903), Chiranthodendraceae A. Gray (1887), Dombeyaceae Kunth ex Desf. (1829) , Durionaceae Cheek (2006), Fremontiaceae J. Agardh, nom. illeg..(1858), Fugosiaceae Martinov, nom. illeg. (1820), Grewiaceae (Dippel) Doweld & Reveal (2005), Helicteraceae J. Agardh (1858), Hermanniaceae Bercht. & J. Presl (1820), Hibiscaceae J. Agardh (1858), Lasiopetalaceae Rchb. (1823), Melochiaceae J. Agardh (1858), Pentapetaceae Spreng. (1826), Philippodendraceae (Endl.) Juss. (1847), Plagianthaceae J. Agardh (1858), Sparmanniaceae J. Agardh, nom. cons.(1858), Sterculiaceae Vent, ex Salisb., nom. cons. (1807), Theobromataceae J. Agardh (1858) and Triplobaceae Raf., (1838)
|Elements of Malvaceae||Other taxa|
|van Royen (1740)||1|
|B. Jussieu (1759)||2|
|A.L. Jussieu, 1774, 1789||4|
|Kunth, 1821, 1882|
|De Candolle, 1824|
|G. Don, 1831|
|Bentham & Hooker, 1865||xx.|
|Masters, 1868, 1875|
|Waton & Dallwitz|
|Thorne & Reveal, 2005|
|1. Geranium (including Pelargonium) (Geraniaceae) is placed in Columniferae.|
1. Stewartia, Camellia and Gordonia (Theaceae or Ternstromiaceae) are placed in Malvaceae; Thea (=Camellia), Tribulus (Zygophyllaceae) and Magnoliaceae in Tiliaceae.
2. Stewartia, Camellia (Theaceae or Ternstromiaceae) and Lagerstroemia (Lythraceae) in Malvaceae; Aceraceae, Hippocastanaceae, Hamamelis (Hamamelidaceae) in Tiliaceae
iii. Malachodendron, Gordonia (Theaceae or Ternstroemiaceae)
iv. Flacourtiacacee, Stewartia (Theaceae or Ternstromiaceae)
xx. Plagiopteraceae, Sphaerosepalaceae
Perhaps the earliest recognition of Malvaceae as a distinct group of plants was by Adrianus van Royen in his Flora Leydensis (1749), who, in an essay at a natural classification of plants divided them into 20 classes, 6 of which are recognisable as modern families. One of these is Columniferae, which in his work incorporates Malvaceae s.s. and Hermannieae. Theobromateae, Grewioideae and Tilioideae (together with Bixa, which was often placed in Tiliaceae during the following century) are to found among a heterogenous grouping to which he gave the name Polyantherae.
Bernard Jussieu in 1759 (reprinted in Antoine Laurent de Jussieu's Genera Plantarum)
¶ The earliest recognition of Malvaceae as a distinct group of plants is difficult to identify. Adrianus van Royen in his Flora Leydensis (1749) incorporates in his Columniferae Malvaceae s. s. and Hermannieae.
The origin of the concept of Malvaceae can be traced back to Adanson and Jussieu, if not to the Columniferae of Linnaeus. Both Adanson and Jussieu may be given as the authority for the names Malvaceae and Tiliaceae.
In his "Philosophia Botanica" (1751) Carl von Linne (Carolus Linnaeus) ... Culminiae, Columniferae
In his "Familles des plantes" (1763) Michel Adanson divides most of the plants now recognised as malvaceous into his 48th (Tilieæ) and 50th (Malveæ) families. The number of genera listed is relatively small,. and consequently many elements of Malvaceae are represented by only a single genus, or are unrepresented, which makes the position of the various elements of Malvaceae in his classification not totally unambiguous. Howerver, his Malveae (pp. 390ff) includes Malvoideae, Bombacoideae (Adansonia, Bombax), Durioneae (Durio), Dombeyoideae (Pentapetes), Byttnerieae (Byttneria) and Hermannieae (Waltheria) as well as a number of genera more recently placed in Theaceae or Ternstromiaceae (Stewartia, Camellia) and Lythraceae (Lagerstroemia). His Tilieæ (pp. 378ff) include Tilioideae and Grewioideae and an assortment of other elements, including Helicteroideae (Helicteres), Theobromeae (Guazuma), Bixaceae, Muntingiaceae and Elaeocarpaceae (Sloanea). as well as a hamamelidaceous plant. Sterculia (as Cavalam) is placed in Tithymalaceae (Tithymali), a family which does not correspond closely to any modern family. Sasali (now considered a synonym of Grewia) and Byttneria are placed in Zizyphaceae (Zizyphi), another family not corresponding closely to any modern family. Hermannia and Melochia (but not Waltheria) are placed in Geraniaceae.
*** Need more of the book to check what he did with Lasiopetaleae and Brownlowioideae ***
In the "Genera Plantarum" (1789) Antoine Laurent de Jussieu also divides the group into two families, the 14th (Malvaceæ) and 19th (Tiliaceae) orders of his 13th class. His Malvaceæ incorporates Malvoideae, Bombacoideae, Helicteroideae, Dombeyoideae, Byttnerieae, Theobromateae and Sterculioideae, as well as a number of genera more recently placed in Theaceae or Ternstroemiaceae (Malachodendron, Gordonia) and Linaceae (Hugonia). His Tiliaceæ incorporates Tilioideae, Grewioideae and Hermannieae, as well as Flacourtiaceae, Elaeocarpaceae and Bixaceae, and Stewartia.
Jussieu expressed doubts as to the incorporation of Hermannieae into Tiliaceae, listing them as Tiliaceæ dubiæ; the remaining genera were given as Tiliaceæ veræ or Genera Tiliaceis affinia.
Jussieu's "Genera Plantarum" was first printed in 1774. I don't know what differences, if any, exist between the 1774 and 1789 printings.
In his "Elementa Botanica" (1790) Necker presents a classification strikingly less modern in aspect than those of Adanson and Jussieu. He divides his plants into 54 groups, all given names ending in -phytum, and although named as genera, corresponding roughly in rank to a family. Three of these groups can be recognised as malavaceous, although in each case contaiing several other genera, some of which are distantly related. Omoplephytum contains genera from Malvoideae and Bombacoidae, and also Dulacia (Olacacae), Mesua (Clusiaceae), Barringtonia (Lecythidaceae) and Camellia, Gordonia and Stewartia (Theaceae or Ternstroemiaceae). Comizophytum contains most genera from Byttnerioideae (including Kleinhovia, but not Byttneria, with Hermannieae as the core of the group), and also 19 other genera, currently placed in several other families. His Dapsilophytum includes a variety of genera from Tilioideae and Grewioideae, other taxa often placed in Tiliaceae by later workers (Muntingia, Bixa, Sloanea, Banara) and a variety of other genera from Capparadaceae, Clusiaceae, Lamiaceae and other families.
His Synathrophytum includes Helicteres, among a considerable number of genera, including members of Magnoliales, Ranunculaceae and Crassulaceae.
In his "Tableau de l'école de botanique du Muséum d'histoire naturelle" (1804) Desfontaines ...
*** May be missing parts of Tiliaceae - get following pages from Gallica ***
¶* Sterculiaceae was introduced by Ventenat in Salisbury's "Parad. Lond" (1807). I have not seen this work, so I am unaware of with which scope the name was introduced.
¶* Byttneriaceae was introduced by R. Brown in his "Botany of Terra Australis" (1814). The restricted scope of this work means that it is not possible to identify what bounds he saw for this, or other families, but it can be seen that his Buttneriaceae incorporated at least Byttnerieae and Lasiopetaleae.
* In the 2nd edition of his "Theorie Elementaire de la Botanique" (1819) Augustin de Candolle divides the malvaceaous plants into Malvaceae, Byttneriaceae, Sterculiaceae and Tiliaceae. Elaeocarpacaceae are not included in Tiliaceae. (Data taken from Lindley's "Vegetable Kingdom" (1847); the composition of his families is not known to me.)
¶ Bombacaceae was introduced by Kunth in Humbert & Bonpland's "Nova genera et species plantarum" (1821). In this work he divides the malvaceous taxa between Malvaceæ, Bombaceæ, Buttneriaceæ and Tiliaceæ. His Bombaceae includes not only Matisieae, Bombacoideae s.s. (and presumably Durioneae), but also Fremontodendreae (Chiranthodendron) and Helictereae (Helicteres). His Malvaceæ is Malvaceae s.s, his Buttneriaceæ corresponds to the Sterculiaceae of other authors, incorporating Sterculioideae, (presumably) Dombeyoideae and Byttnerioideae, and his Tiliaceæ Brownlowioideae, Grewioideae and Tilioideae.
¶* In his "Mag. Aesth. Bot." (1823) Reichenbach introduced the family Lasiopetalaceae (Lasiopetaleae). I have not seen this, so I am unaware of with which scope the name was introduced
In his "Prodromus Systematis Naturalis Regni Vegetabilis" (1824) Augustin de Candolle divides the malvaceous taxa into four families in much the same way as Kunth. However, Plagianthus and Montezuma (a segregate of Thespesia) are placed in Bombacacae rather than Malvaceae, and Goethea (a segregate of Pavonia) in Byttneriaceae rather than Malvacee. His Tiliaceae includes Muntingiaceae and Gyrostemonacaeae.
In his "Catalogus plantarum horti regii parisiensis, cum annotationibus de plantis novis aut minus cognitis" (1829) Desfontaines has a classification similar to the De Candolle's but segregates, separately, Hermannieae and Dombeyoideae, as Hermanieae and Dombeyeae. Muntingia and Gyrostemonaceae are not included in Tiliaceae, but Bixaceae and Flacourtia are.
* In his "Ordines Naturales Plantarum" (1830) Bartling divides the malvaceous families into the same six families (Malvaceae, Dombeyaceae, Hermanniaceae, Byttneriaceae, Sterculiaceae and Tiliaceae) as Desfontaines above, but excludes Bixaceae from Tiliaceae. (Data taken from Lindley's "Vegetable Kingdom" (1847); the composition of his families is not known to me.)
In his "General History of the Dichlamydeous Plants" (1831) G. Don divides the malvaceous families into 5 families, in a manner similar to that of De Candolle, but with Sterculioideae (and Reevesia) segregated as Sterculiaceae. His Tiliaceae includes Gyrostemonaceae.
* In his "Nixus Plantarum" (1833) Lindley divideds the malvaceaeous taxa into the three families Malvaceae, Sterculiaceae and Tiliaceae. Elaeocarpaceae and Bixaceae are excluded from Tiliaceae. (Data taken from Lindley's "Vegetable Kingdom" (1847); the composition of his families here is not known to me.)
In his "Histoire Naturelle des Végétaux" (1834) Spach divides the Columniferae into the 5 families Malvaceae, Sterculiaceae, Dombeyaceae, Hermanniaceae and Byttneriaceae (I haven't seen the volume which covers Tiliaceae). He divides Malvaceae into two tribes, the first (Malveae) corresponding to Malvaceae s.s., and the second (Bombaceae) incorportating Bombacoideae (including Fremontodendreae) and Helicteroideae. (Helicteres apetala is misplaced in Sterculia, and Reevesia treated as an anomalous genus in Sterculiaceae.) Dombeyaceae corresponds to Dombeyoideae, with the addition of Kydia. Hermanniaceae corresponds to Hermannieae, and Byttneriaceae to the remainder of Byttnerioideae.
(1835) Martius ...
In his "Repertorium botanices systematicae" (1842) Walpers divides the malvaceae taxa into the 5 families Malvaceae, Sterculiaceae, Byttneriaceae, Philippodendraceae and Tiliaceae. His Sterculiaceae includes Matisieae, Bombacoideae, Helicteroideae and Sterculioideae; his Byttneriaceae Dombeyoideae and Byttnerioideae; his Philippodendraceae Plagianthus regius; and his Tiliaceae includes Brownlowioideae, Grewioideae and Tilioideae, as well as Muntingiaceae and Elaeocarpaceae. Plagianthus and related genera are placed in subtribe Myrodieae of tribe Helictereae in Sterculiaceae.
In his "Vegetable Kingdom" (1847) Lindley ...
¶ In his "Theoria systematis plantarum" (1858) Agardh divides malvaceous plants into Sparmanniaceae, Byttneriaceae (Büttneriaceae), Sterculiaceae, Helicteraceae (Helictereae), Theobromataceae (Theobromeae), Fremontiaceae (Fremontieae), Bombacaceae (Bombaceae), Tiliaceae, Lasiopetalaceae (Lasiopetaleae), Melochiaceae (Melochieae), Plagianthaceae (Plagiantheae), Malvaceae and Hibiscaceae (Hibiscseae). Of these families Fremontiaceae, Helicteraceae, Hibiscaceae, Melochiaceae, Plagianthaceae, Sparmanniaceae and Theobromataceae are original to this work.
In Oliver's "Flora of Tropical Africa" (1868) and Hooker's "Flora of British India" (1875), Masters divides malvaceous plants into the three families Malvaceæ, Sterculiacæ and Tiliaceae. Bombacoideae and Durioneae are placed in Malvaceæ. Elaeocarpaceae is included in Tiliaceæ
In his "A Handbook of Systematic Botany" (1878)) Eugenius Warming divides the Columniferae into Malvaceae (Malvoidae, Bombacoideae and Durioneae), Sterculiaceae (Byttneroideae, Dombeyoideae, Helictereae and Sterculioideae) and Tiliaceae (Brownlowioideae, Grewioideae, Tilioideae, and also Elaeocarpaceae).
¶* In "Proceedings of the American Academy of Arts and Sciences 22" (1887) Asa Gray introduces a family Chiranthodendraceae (Chiranthodendreae), corresponding to Fremontodendreae.
¶* In Post & Kuntze's "Lex. Gen. Phan" (1903) Augier's Cacoaceae is introduced. I have not seen this work, and do not know the scope of this taxon, nor how the remainding malvaceous plants are classified.
¶* In his "Prosyllabus Tracheophytorum, Tentamen systematis plantarum vascularium (Tracheophyta)" (2001) Doweld introduced a family Berryaceae. I have not seen this book, and do not know the scope of this family, or the division of the rest of Malvaceae in his classification.
¶* In "Botanical Review (Lancaster) 71: 100" (2005) Doweld and Reveal introduce a family Grewiaceae. I have not seen this journal and do not know the division of the rest of Malvaceae in their classification. Note the independent introduction of Grewiaceae (equivalent to Grewioideae) by Thorne. Another recent book "Plant Systematics: An Integrated Approach" (date and author unknown, probably Gurcharan Singh2004) has two families - Malvaceae and Grewiaceae.
¶ first use of family
* omitted from table
Malvaceae Info (Home)
© 2006, 2007 Stewart Robert Hinsley
The Angiosperm Phylogeny Group (APG) started with ordinal relationships for angiosperm families and later came to concentrate more on issues surrounding family delimitation, generally taking a conservative approach (favouring lumping over splitting when choices were otherwise fairly evenly balanced). The history of the APG approach to classification is elaborated here, and long-term trends in taxon size are also discussed. Twenty-three questions relating to family delimitation of angiosperms (21) and ferns (two) were identified, and an open, web-based survey was conducted to determine which solutions were preferred by the botanical community. There were a total of 441 responses from 42 countries. The full survey results are presented and discussed. In general, among respondents, there was a good balance between lumping and splitting, although a much smaller percentage self-identified as splitters. Recommendations to APG for a set of specific family delimitation issues in angiosperms and ferns are made. We followed the majority vote in all but four cases and recommended ten splitting and 13 lumping options. We found that decisions were made by respondents on a case-by-case basis. Many respondents favoured lumping as a means of gaining stability by leaving circumscription issues at the lower taxonomic levels, which are of more concern to specialists (i.e. at subfamilial and tribal levels). We estimate that APG IV will not differ substantially from APG III.
In an effort to bring about another revision of the Angiosperm Phylogeny Group (APG) classification of the orders and families of angiosperms (to be called APG IV), a form of impasse has been reached. A consensus of a large number of researchers was behind the earlier versions of APG (APG, 1998; APG II, 2003; APG III, 2009; see the list of consulted experts provided in each) and, following the Backlund & Bremer (1998) principles, most divergences of opinion were relatively easily resolved. Circumscription issues up to 2003 (APG, 1998; APG II, 2003) had mostly to do with the embedding within a larger well-known family of small sets of genera split off as a satellite family by some authors, and, in such cases, a consensus on how to proceed was easily reached: monophyly presented the authors with two choices and, given that most of the larger units had long histories of recognition, to maximize stability only one of these options was clearly viable. Beginning with APG II (2003), the issue of choosing between narrower or broader circumscriptions had begun to become a focus, and this issue assumed even larger proportions by the time of APG III (2009). The other coincident phenomenon was that the remaining ‘problems’ requiring taxonomic solutions were for newly discovered groups or those for which there has never been a consensus of taxonomic opinion. Examples of the former are the unfamiliar families of orders Huerteales and Crossosomatales and of the latter some well-known and long-recognized families, such as Dioscoreaceae, Lamiaceae, Scrophulariaceae, Portulacaceae, Aizoaceae and Santalaceae, for which an accumulating body of research had demonstrated that some degree of major reorganization was required. Such problems have little to do with the monophyly issue per se (although this still forms the underlying framework in which these various options are posed), and their solutions are almost entirely concerned with lumping versus splitting approaches, i.e. where in the cladogram the new lines are to be drawn. The use of the Backlund & Bremer (1998) or Vences et al. (2013) principles in these cases does not make a clear decision possible. Many have argued that such decisions are arbitrary (with which to some extent we agree; Barraclough & Humphreys, 2015) and therefore without scientific merit (a view with which we do not agree; see below). We decided that one way to take matters forward and to develop a framework for how to view a long list of problems on the agenda for APG IV was to undertake an online survey of taxonomic opinion among our peers about the issue of splitting versus lumping as solutions to these problems. In the longer term, it is use that determines whether a given taxonomic scheme is taken up, and so by determining preferences a priori, we hoped to make APG IV more closely reflect users’ wishes. In a general sense, APG has been widely accepted, especially by our peers in other areas of science (e.g. by GenBank), but also by influential organizations, such as the Royal Horticultural Society through its annually produced Plant Finder (Cubey, Edwards & Lancaster, 2014), and indeed we hope to make APG as useful as possible to a wide variety of users of plant names in commerce, but especially, although not exclusively, including our taxonomic peers. Before presenting these results, we provide further background about the APG process up to the present.
The original APG classification (APG, 1998) ushered in a new way of performing classification and was the first to incorporate large amounts of DNA sequence data into a re-classification for a major group of organisms. Sibley & Ahlquist (1990) first used DNA information to revise the classification of birds (in this case, DNA–DNA hybridization kinetic data), but this data type was difficult to interpret and use in the reconstruction of phylogenetic trees. Nonetheless, the method of constructing a consensus classification and basing it on an explicitly analysed set of data, rather than on an intuitive interpretation of the available data, set this classification of birds and the APG classification apart from all previous such efforts. APG (1998) was also noteworthy for the fact that it was not authored by one authority, but rather was based on an attempt to reach an informal consensus among many researchers (a brief description of this process was included in APG, 1998). The primary focus of this first version of the APG classification was which families belonged to which orders; given the limits of the available data at that time, circumscription of families was only a minor issue. The original driving force of APG was communication; the authors wished to avoid the confusion caused by the same clades of families being called different names by the various authors working in the field of molecular systematics. For example, D. Soltis and colleagues, working on Saxifragaceae and related families, and Chase and colleagues, working on Malpighiaceae and related families, both members of Rosidae sensu Cronquist (1981) but in different orders, found that not only were their respective taxa emerging as interdigitated, but they were also mixed up in clades with members of families placed in Asteridae sensu Cronquist, which were being sequenced by Olmstead, B. and K. Bremer and colleagues. Of course, few representatives of each family were being included in these studies, given the limitations imposed by sequencing technologies at that time. Thus, the original APG circumscriptions of families followed those of previous angiosperm classifications, such as Cronquist (1981), Thorne (1992) and Takhtajan (1980). Takhtajan (1997) followed a somewhat different course and split many families, only to re-associate them as orders, a form of taxonomic inflation, but clustering of sets of genera was still similar, albeit at a different level in the hierarchy in Takhtajan versus Cronquist or Thorne. Disagreements among the APG authors did occur, but they were centred on what to call one particular clade of families, e.g. Violales, Euphorbiales, Linales or Malpighiales (all these names had been applied to the clade now called Malpighiales in APG, and all were potential candidates, given that priority does not apply at this hierarchical level). All broad studies produced similar sets of families forming clades, and what was required of these authors was to decide which particular ordinal names should be applied to these.
APG II and APG III
By the time of APG II (2003), the issue of family circumscriptions was becoming a more important focus, and the first update of APG (APG II, 2003) included the so-called ‘bracketed’ system, whereby an author could use either a broader or narrower circumscription of several families and still claim to be following APG. This ‘flexibility’ proved to be highly unpopular based on feedback received by the APG ‘compilers’ (B. Bremer, K. Bremer, M. Chase, J. Reveal, D. Soltis, P. Soltis and P. Stevens), which is why its use was abandoned in APG III (2009). In the process of eliminating the bracketed options, in the great majority of cases APG III adopted the broader of the two optional circumscriptions, which seemed to be the most popular choice, again based on informal feedback the compilers received from many sources.
Prior to the publication of APG III, two consultative processes were initiated. Because of the problems raised by the ‘bracketed’ system and work published after APG II, David Mabberley conducted a series of workshops in the early 2000s involving staff at several major herbaria in Australia, Europe and the USA in an effort largely intended to gain consensus. In consequence, his system, published in Mabberley's plant-book (3rd edition, 2008, pp. 927–938), largely coincided with what was to become APG III. In 2008, faced with upcoming physical reorganizations/moves of several major European herbaria, the Vascular Plant Classification Committee (Wearn et al., 2013) was established under the joint chairmanship of Mark Chase and David Mabberley, then both at Kew, with the idea of creating a consensus about which classification of plants should be followed for organizing the collections at the following member herbaria: Natural History Museum, London (BM), Royal Botanic Garden Edinburgh (E), Conservatoire et Jardin Botaniques, Genève (G), Royal Botanic Gardens, Kew (K), Nationaal Herbarium Nederland (L, U and WAG, now part of NCB Naturalis) and the Muséum National d'Histoire Naturelle, Paris (P). Input was also sought from other international authorities (e.g. in Australia and North America), and an agreement was reached in early 2009 to follow an unmodified APG III linear system (Haston et al., 2009), thereby again providing support for the mostly broader circumscriptions in that classification. Criticisms of this ‘lumping’ approach have been published and aired at major international symposia (Heywood et al., 2007; Martínez-Azorín et al., 2013), but, in general, the APG approach of recognizing fewer and more broadly circumscribed families has received broad support.
Size/number of families
The most vexing issue still facing APG is how best to deal with the issue of family circumscriptions. The advent of DNA-based classifications seems to have led to an increase in the size of genera (Humphreys & Linder, 2009). Many botanists consider the issue of splitting versus lumping to be inconsequential and of no real concern to a scientist, given that classification at these levels is an artificial construct (e.g. Atran, 1990; Barraclough & Humphreys, 2015) and cannot be falsified. What really is the difference between recognizing a given set of genera as subfamilies or families? In many contexts this is true, but this response does not take into consideration how non-taxonomists and students utilize classifications. Users and students generally focus on a small number of categories, mostly species/genus and families (Vences et al., 2013), and, as a result, they are oblivious to and rarely use supra- and infra-familial categories. Therefore, there is a consequence of using a narrow family concept and recognizing more, smaller families: students and other users learn and use a smaller percentage of the recognized families, with the result that they are confronted by commonly encountered but unfamiliar families as they continue their careers in botany. For example, Agapanthus L'Hér., if treated as the sole genus of Agapanthaceae (APG, 1998), will never be taught in a plant taxonomy course, but it is a common horticultural subject. Students will never learn about Agapanthaceae, but, if Agapanthus is treated as a member of Amaryllidaceae sensu APG III, they will immediately recognize the family and understand where it fits into the system. The more complicated the system, the more daunting it is to gain a foothold in the field as a student, and the more unlikely it is that non-specialist users will be able to use it effectively. For this practical reason, one focus of APG has been the reduction of the number of families (e.g. lumping, as is evident in APG III versus APG II); this keeps the system simpler and makes it more approachable.
Recognition of families
Lumping has been criticized as problematic if the larger families created are so heterogeneous that they lack obvious synapomorphies (see comments in survey results below). Thus, the other major issue that accompanies arguments over lumping versus splitting has been that of diagnosability. If all families were easily recognized by clear synapomorphies, it would make such a classification easier to remember and use, and there has long been a tradition of trying to make sure that whatever sets of genera we recognize as families have some obvious characters (the principle of ‘maximizing ease of identification’ in Backlund & Bremer, 1998, and ‘phenotypic diagnosability’ in Vences et al., 2013). In several cases, this has clearly become a challenging objective. There are some angiosperm families that have a long history of recognition, going back as far as the ancient Greeks or at least to the time of John Ray (17th century); for example, daisies, palms, grasses, mints, legumes, etc. have a clear set of synapomorphies, which have been recognized for a long time. An earlier example is that of Morison (1672), whose treatment of Apiaceae (as Umbelliferae) was the first western taxonomic monograph. These clear-cut groups stand out starkly in comparison with the situation in ferns and some groups of angiosperms, such as those referred to in the past as Saxifragaceae s.l. and Liliaceae s.l. Once we entered the molecular era, it became clear that the consensus family concept (in operation at least since the time of Jussieu, 1789) was not easily applicable to all groups of angiosperms in the same manner as it was to bromeliads, daisies, gentians, grasses, legumes, orchids, palms and many others with a long history of recognition, and which coincided with clades in molecular studies. That molecular analyses were so much in agreement with long-held circumscriptions of families is surprising in some contexts, given that classification by consensus has long been the norm in botany (Stevens, 1994) and consensus in this case was largely based on intuition (but of many well-informed botanists).
In the problematic groups, the issue of lumping versus splitting becomes important. If there are obvious synapomorphies, these often apply to only some of the genera concerned or are developmental or phytochemical characters, leaving the impression that these are somehow not equivalent to the well-circumscribed larger clades that have a long history of recognition as families (e.g. orchids, daisies, legumes, etc.). Chase, Fay & Savolainen (2000) estimated that DNA studies confirmed the family limits of c. 87% of the families recognized by Cronquist (1981); however, in the remaining 13% (c. 50 or so families), there are some major problems in circumscription (e.g. Euphorbiaceae, Flacourtiaceae, Liliaceae, Portulacaceae, Santalaceae). Of course, Cronquist was merely the last proponent of a series of family circumscriptions that goes back to Jussieu (1789