Pre-Darwinian

The type genus, Lilium, from which the name of the family was derived, was originally formally described by Carl Linnaeus in 1753, with seven species.^[3] He placed Lilium within the Hexandria Monogynia (six stamens, one carpel) in his sexual classification in the Species Plantarum.^[4] The family Liliaceae was first described by Michel Adanson in 1763,^[5][6] but formally named by Antoine Laurent de Jussieu in 1789.^[1][7] Adanson described eight subfamilies with 78 genera, however the subfamily he described as Lis (lilies) had seven genera (Uvularia, Mithridatium, Mendoni, Lilium, Fritillaria, Imperialis — now part of Fritillaria — and Tulipa) of which four are in the modern genus.^[8] Jussieu placed these into the ordo, Lilia in the classis, Stamina Perigyna of the Monocotyledones (monocots), with eight genera (Tulipa, Erythronioum, Methonica, Uvularia, Fritillaria, Imperialis, Lilium, Yucca) only four of which remain in the family. He defined Lilia as "calix" (perianth) of six equal coloured parts, six stamens, a superior ovary, single style, and trilocular capsule. The term ordo at that time was closer to what we now understand as family, rather than order.^[9][10] Although Jussieu used the Latin "lilia" in his Genera Plantarum, elsewhere he used the French "liliacées",^[11] as had Adanson. The word "Liliaceae" was soon widely used by botanists such as Samuel Frederick Gray,^[12] John Lindley,^[13] and Pierre-Joseph Redouté^[14] in the early nineteenth century.

Gray (1821) provided the first description of Jussieu's scheme in English, identifying two genera occurring in Britain (Tulipa, Fritillaria), distinguished by the absence or presence of basal nectaries. His key used the presence of six equal stamens, a single style, a simple petaloid (undifferentiated tepals resembling petals) perianth and a trilocular capsule with flat seeds to identify the family.^[15] Although Augustin De Candolle (1813) had not explicitly described the Liliaceae, his overall classification scheme influenced many later writers including Gray.^[16] In this scheme, ^[17] the Liliaceae were considered a family within those vascular plants (Vasculares) whose vascular bundles were thought to arise from within (Endogènes, endogenous), a term he preferred to Monocotylédonés. Jussieu's Monocotyledones became the Phanérogames (Phenogamae in Gray), meaning "visible seed", hence Endogenæ phanerogamæ.^[18] Candolle also instituted the concept of ordered ranks, based on classes, subclasses, familles (Latin: ordines naturales) and tribus (tribes),^[10] subdividing the Liliaceae.

Lindley was the first post-Linnaean English systematist, publishing his work in 1830,^[19] and following the reasoning of Jussieu he used the term tribe to describe the Liliaceae as a division of the hexapetaloid monocots, characterised by a superior ovary, highly developed perianth, inward turning anthers, a trilocular polyspermous capsule and seeds with a soft spongy coat. He offered seven genera as examples (Erythronium, Lilium, Calochortus, Blandfordia, Polianthes, Hemerocallis and Funkia). By 1846, in his final work, he refined and greatly expanded his taxonomy, favouring the term Alliances of Endogens over monocots as a class, of which there were eleven. Of these "alliances", the Liliales consisted of four Orders (families) including Liliaceae, which he referred to as lilyworts in the vernacular, with 133 genera and 1200 species.^[20] In this work he unhappily acknowledged the confusing array of different approaches to the classification of the Liliaceae, the lack of a clear definition, and the great diversity in the circumscription of the order, which had expanded vastly, with many subdivisions. As he saw it, the Liliaceae had already become a ("catch-all") grouping,^[21] being "everything that does not belong to the other parts of the Lilial Alliance", but expressed hope that the future would reveal some characteristic that would group them better.^[22] In other words, he foresaw that Liliaceae would come to be regarded as paraphyletic. By the time of the next major British classification, that of Bentham and Hooker in 1883 (published in Latin) several of Lindley's other families had been absorbed into the Liliaceae.^[23] This was the last major classification using the "natural" or pre-evolutionary approach to classification, based on characteristics selected a posteriori in order to group together taxa that have the greatest number of shared characteristics. This approach, also referred to as polythetic was superseded by ones based on an understanding of the acquisition of characteristics through evolution, referred to as phyletic.^[24][25][26]

Post-Darwinian

Although Charles Darwin's Origin of Species (1859) preceded Bentham and Hooker's publication, the latter project was commenced much earlier and George Bentham was initially sceptical of Darwinism.^[24] The new phyletic approach changed the way that taxonomists considered plant classification, incorporating evolutionary information into their schemata, but this did little to further define the circumscription of Liliaceae.^[27] The major works in the late nineteenth and early twentieth century employing this approach were in the German literature, the Eichler (1875–1886), Engler and Prantl (1886–1924) and Wettstein (1901–1935) systems. These placed the Liliaceae into one of the major subdivisions of the monocotyledons, the Liliiflorae.^{[28][29][30][31][32]} In the English literature, Charles Bessey (1915) followed Adolf Engler in defining Liliaceae as "Pistil mostly 3-celled; stamens 6; perianth of two similar whorls, each of three similar leaves", although placing the Liliales in a novel subclass of monocots, the Strobiloideae,^[33] while from John Hutchinson (1959) onwards the Liliaceae were treated as part of the Liliales (see Table 1).

Over time the Liliaceae became increasingly broadly, and somewhat arbitrarily, defined as all species of plants with six tepals and a superior ovary. They eventually came to encompass about 300 genera and 4,500 species, within the order Liliales in the scheme of Arthur Cronquist (1981).^[34] Cronquist was a "lumper" preferring a small number of very large groupings and his classification of the Liliaceae represented the greatest expansion of the family to date. Cronquist placed most flowering petaloid monocots with six stamens in this very broad (and clearly polyphyletic) family, hence the alternative name lilioid monocots.^[35] He rejected the importance of ovary position and thus included with the Liliaceae with their superior ovary (hypogynous), the Amaryllidaceae, some species of which had an inferior ovary (epigynous), and which others separated into a distinct family.^[36][37][38] The Liliaceae were one of the major families in the Cronquist system which included 22 families in addition to Liliaceae in the more restricted sense (sensu stricto, s.s.) used by others.^[39][40] Later more conventional schemes include the system of Robert F. Thorne^[38] and that of Armen Takhtajan^[41] which characterised the family as petaloid monocots, characterised by showy flowers with tepals and without starch in the endosperm.

Deconstructing Liliaceae

Other botanists in the twentieth century echoed Lindley's concerns about the lack of a clearly defined grouping for Liliaceae. The earliest of these was Johannes Paulus Lotsy (1911),^[42] building on Wettstein's work. Lotsy suggested four separate families, Liliaceae, Alliaceae, Agapanthaceae and Gilliesiaceae, within the Liliifloren. This recognised the major groupings that would later be transferred to Amaryllidaceae as subfamilies Allioideae and Agapanthoideae, with Gilliesieae as a tribe within the Allioideae. This approach was later followed by Herbert Huber in 1969.^[43] These various proposals to separate small groups of genera into more homogeneous families made little impact until Rolf Dahlgren (1985), following Huber's lead,^[44] developed a system incorporating new information, including synapomorphic characters (i.e., shared characters believed to have evolved from a common ancestor).^[35] While Cronquist was a "lumper", Dahlgren was a "splitter", preferring a larger number of more homogeneous groupings. Where Cronquist saw one family, Dahlgren saw forty distributed over three orders (predominantly Liliales and Asparagales), reducing Liliaceae to ten genera (see Table 3).^[45][46] Over the 1980s, in the context of a more general review of the classification of angiosperms, the Liliaceae were subjected to more intense scrutiny. By the end of that decade, the Royal Botanic Gardens at Kew, the British Museum of Natural History and the Edinburgh Botanical Gardens formed a committee to examine the possibility of separating the family into smaller taxa, at least for the purpose of organizing their herbaria. That committee finally recommended that 24 new families be created in the place of the original broad Liliaceae, largely by elevating subfamilies to the rank of separate families.^[40][47]

The 1990s saw considerable progress in plant phylogenetics and cladistic theory, enabling a phylogenetic tree to be constructed for the flowering plants.^[48] The establishment of major new clades necessitated a departure from the older but widely used classifications such as Cronquist and Thorne, based largely on morphology rather than genetic data. These developments complicated discussions on plant evolution and necessitated a major taxonomic restructuring.^[49][50] rbcL gene sequencing and cladistic analysis of monocots in 1995 had redefined the Liliales order^[51] out of four original morphological orders sensu Dahlgren. The largest clade within this redefined Liliales, christened the "core Liliales" representing a now much reduced Liliaceae, had all previously been included in the Liliales, and included three taxonomic groups: (i) Liliaceae sensu Dahlgren,^[52] but also both the (ii) Calochortaceae as defined by Minoro Tamura (sensu Tamura)^[53] and (iii) Clintonia-Medeola which had been included in Liliaceae sensu Tamura^[53] (see Table 3).^[54]

This newly, more narrowly (sensu stricto, s.s.) circumscribed Liliaceae, corresponded to the emerging circumscription of the family in the Angiosperm Phylogeny Group system (1998).^[54] While this also corresponded most closely with Dahlgren's circumscription relative to the older much broader (sensu lato, s.l.) constructions, it gave rise to some potentially confusing terminology. Compared to the very broad historical s.l. construction of Liliaceae, the Dahlgren, Tamura and APG constructions were much narrower. These have variously been referred to as "core Liliales" and sensu APG^[54] for the broadest construction, while the intervening schemes of Tamura and Takhtajan (who was also a "splitter")^[55] which are narrower have been referred to as Liliaceae s.s., sensu Dahlgren, and sensu Tamura. Of these the narrowest circumscription is that of Dahlgren, including only Lilieae s.l..^[54] In modern terminology, while "core Liliales" represents Liliaceae sensu APG, Liliaceae sensu Tamura corresponds to Lilioideae s.l., and sensu Dahlgren with Lilieae s.l. or Lilioideae s.s.. Calochortoideae and Streptopoideae sensu APG represent and Calochortaceae sensu Tamura, and Clintonia plus Medeola (Medeoleae) Medeoloideae sensu Tamura (see Table 3).^[2]

Liliaceae

The diversity of characteristics complicates description of Liliaceae morphology, and confused taxonomic classification for centuries. The diversity is also of considerable evolutionary significance (see Evolution).^[54] The family Liliaceae are characterised as monocotyledonous, perennial, herbaceous, bulbous (or rhizomatous in the case of Medeoleae)^[97] flowering plants with simple trichomes (root hairs) and contractile roots.^[108] The flowers may be arranged along the stem, developing from the base, or as a single flower at the tip of the stem, or as a cluster of flowers. They contain both male (androecium) and female (gynoecium) characteristics and are symmetric radially, but sometimes as a mirror image. Most flowers are large and colourful, except for Medeoleae. Both the petals and sepals are usually similar and appear as two concentric groups (whorls) of "petals", that are often striped or multi-coloured, and produce nectar at their bases. The stamens are usually in two groups of three (trimerous) and the pollen has a single groove (monosulcate). The ovary is superior, i.e. placed above the attachment of the other parts. There are three fused carpels (syncarpous) with one to three chambers (locules), a single style and a three-lobed stigma. The embryo sac is of the Fritillaria type. The fruit is generally a wind-dispersed capsule, but occasionally a berry (Medeoleae) which is dispersed by animals. The leaves are generally simple and elongated with veins parallel to the edges, arranged singly and alternating on the stem, but may form a rosette at the base of the stem.

Subclades

(See Table 3). The ten genera (two genera of Table 3 are subsumed into other genera) of the subfamily Lilioideae are characterised by contractile bulbs and roots and a megagametophyte (embryo-sac) of the Fritillaria-type with four megaspores. Within the Lilioideae, the eight genera considered as Liliaceae by Dahlgren (sensu Dahlgren), that is Lilieae s.l., are characterised by loculicidal capsules and a basic chromosome number x=12. Within this clade, Lilieae s.s. are characterised by papillose tepals (with the exception of Fritillaria) and numerous fleshy bulb-scales as well as a morphologically distinct karyotype with two long metacentric chromosomes and 10 telocentrics of medium length.^[109] The two genera within Tulipeae are distinguished by pseudo-basifixed anthers and single bulb scales. The two genera of Medeoleae are distinguished by having rhizomes instead of bulbs and berries instead of capsules, and a very unusual form of vesicular-arbuscular mycorrhizae.^[54]

The five genera constituting the Streptopoideae and Calochortoideae subfamilies form another distinct group, previously characterised under the Calochortoideae alone. These have creeping rhizomes, styles divided at their apices, and an embryo-sac of the Polygonum-type with a simple megaspore and triploid endosperm. At times, these genera were considered as a separate family (Calochortaceae; e.g. Tamura) or even placed in the more heterogeneous Uvulariaceae sensu Dahlgren. However most of the latter had low morphological similarity to the Liliaceae, and Uvularia and Disporum are now classified in the Colchicaceae. Disporum contained both Asian and North American species which had always been distinguishable. Following molecular analysis, the North American species were restored to the genus Prosartes and retained in Liliaceae, subfamily Streptopoideae, while the Asian species were moved to Colchicaceae.^{[54][110][111]}

Suprageneric subdivisions

Due to the diversity of the originally broadly defined Liliaceae (s.l.), many attempts have been made to form supageneric classification systems, organizing the genera into subfamilies, tribes, or other suprageneric taxa (taxonomic groupings between genus and family).^[94] By 1813, Candolle recognised five subdivisions which he called tribes (Asparagées, Trilliacées, Asphodelées, Bromeliées, Tulipacées),^[17] all of which Jussieu had made separate families, with the exception of Tulipa, which was a genus within the Liliaceae. By 1845, John Lindley observed that the family had become extremely diverse, ill-defined and unstable, not only by its overall circumscription, but also by its subdivisions. For the 133 genera he included, he described eleven suborders.^[20] By the 1870s, as Baker describes in his revision of the family, the taxonomy of Liliaceae had become vast and complicated. His approach was to divide the family into eight tribes.^[112][113]

In 1879, a revision of the North American Liliaceae by Sereno Watson described sixteen tribes, of which the Lilieae correspond most closely to current concepts of the family,^[114] Bentham and Hooker described twenty tribes in 1883, and Engler and Prantl in their extensive description of the Liliaceae in 1899 identified 31 tribes distributed over 11 subfamilies, with Tulipeae and Alieae representing the modern family.^[115] In 1936, Franz Buxbaum undertook a major revision of the Liliaceae and among others described the subfamily Lilioideae with three tribes: Lloydieae (Gagea, Lloydia and Szechenya and Giradiella — both now included in Lloydia), Tulipeae (Erythronium, Tulipa and Eduardoregalia — now part of Tulipa) and Lilieae (Korolkowia, Fritillaria, Notholirion, Cardiocrinum, Nomocharis and Lilium).^{[116][117][118]} Hutchinson included most of these genera within the tribe Tulipeae.^[91] The complex rearrangements of the various genera, tribes and subfamilies over a 30-year period from 1985, discussed by Peruzzi and colleagues (2009),^[94] are partly summarised in Table 2 below.

Classifications published since the use of molecular phylogenetics have taken a narrower view of the Liliaceae (s.s.). In 1998, Tamura considered Calochortus sufficiently distinct to elevate the subfamily Calochortoideae to family status as Calochortaceae,^[97][53] resulting in the term Liliaceae sensu Tamura to indicate Liliaceae without the Calochortoideae. In 2009, Takhtajan used an even narrower definition (see Table 2 & Table 3 below).^[87]

Despite now having established a taxonomic grouping for the family Liliaceae that is genetically monophyletic, compared to the prior longstanding polyphyletic assemblages under this name,^[51][119] the morphology remains diverse,^[97] and there exists within the Liliaceae clade a number of subclades. There appeared to be two major clades, the first and largest of these consisted of three subclades: Clintonia—Medeola—Gagea, Lilium Fritillaria—Notholirion, and Tulipa—Erythronium. The second smaller clade, Streptopus—Tricyrtis contained elements of Dahlgren's Uvulariaceae. The position of Calochortus remained problematic, being considered a sister clade to Liliaceae, as treated by Tamura, but further analysis suggested it was in fact sister to Tricyrtis, although it is now considered separate once again (see Table 3).^{[63][54][93][120][121]}

Also enigmatic were Clintonia, Medeola, Scoliopus, and Tricyrtis. Clintonia, with a disjunct distribution involving East Asia and North America, and the closely related Medeola form a subclades and are now considered a separate tribe (Medeoleae) within the Lilioideae, although at different times they have been considered a separate subfamily (Medeoloideae) or family (Medeolaceae). Sequencing of the rbcL and matK chloroplast genes established monophyly for Clintonia, but with separate clades corresponding to the two areas of distribution.^[122] The Angiosperm Phylogeny Website (APWeb)^[2] includes four of Takhtajan's families in Liliaceae, recognizing three subfamilies, one of which is divided into two tribes and referred to as Liliaceae sensu APG III.^[123]

Genera