Ipomoea as constituted by Linnaeus was based on Ipomoea pes-tigridis L. and contained various elements, including I. quamoclit and I. coccinea (Quamoclit Clade, page 556), I. triloba and I. lacunosa (Batatas Clade, page 387), I. violacea, I. alba and I. carolina as well as species of Merremia Dennst. ex Endl. and Jacquemontia Choisy and even a species of Hydrophyllaceae, I. nyctelea L. (=Ellisia nyctelea (L.) L.). It was not clearly defined and several species since treated as belonging to Ipomoea were placed in Convolvulus L. by Linnaeus including I. purpurea and I. pes-caprae.

Jacquin, Vahl, Willdenow and others of Linnaeus’ successors in the later part of the 18^th century continued placing species of Convolvulaceae rather arbitrarily in either Convolvulus L. or Ipomoea. Only Cavanilles’ placements came close to coinciding with a modern concept of I.omoea. Some authors, like Desrousseaux (1792), maintained a wide concept of Convolvulus that included all species of Ipomoea and it was only in 1810 that a clear distinction between the two genera, based on stigma morphology, was established by Robert Brown (1810: 484). He contrasted the 2–3-lobed, capitate stigma of Ipomoea with the two filiform stigmas of Convolvulus. Brown recognized the ovary of Ipomoea as being 2–3 locular but made no attempt to subdivide the genus based on the number of ovary cells. Although Roemer and Schultes (1819) followed Brown’s classification, a wide circumscription of Convolvulus remained current for some time. Both Kunth (1819) and Sprengel (1824 1827) included Ipomoea within Convolvulus and it was not until the various publications of Choisy (1834, 1838, 1845) that Ipomoea was permanently separated from Convolvulus.

Choisy (1834, 1845) subdivided Ipomoea s.l. into several genera based on a series of ovary and fruit characters. He recognized a tribe Argyreieae Choisy comprising a heterogeneous group of genera including Argyreia, Rivea, Legendrea (=Turbina) and Marcellia on the basis of their having indehiscent fruits. The tribe Convolvuleae Choisy, in contrast, was characterized by having dehiscent fruiting capsules. In this second group, Choisy recognized Quamoclit, Mina, Batatas, Pharbitis, Calonyction, Exogonium and Lepistemon as distinct from but related to Ipomoea based on characters of the ovary and corolla. Quamoclit was recognized as distinct because of the 4-locular ovary, each cell with a single seed. Mina was separated from Quamoclit because of the suburceolate corolla shape. Together these two genera comprise what we recognize as the Quamoclit Clade (page 556). Pharbitis was separated on the basis of having a 3-lobed stigma and 3-locular ovary, each cell with two seeds, this genus constituting the Pharbitis Clade (page 430). Choisy’s Batatas was vaguely defined and is very heterogeneous comprising many extraneous elements besides I. batatas and I. triloba. Calonyction and Exogonium were separated from Ipomoea on the basis of their corolla, large, showy, white or pale lilac in the case of Calonyction but merely tubular in the case of Exogonium. A small clade of species of which I. alba is the best known more or less coincides with Choisy’s Calonyction, which was redefined and extended by Hallier (1897b). Exogonium was accepted by House (1908a) and other authors but is very heterogeneous and therefore polyphyletic, so bearing no clear relationship to the clades recognized in our molecular studies. Lepistemon was separated because of the large scales at the base of the stamens, a character that sometimes appears elsewhere in the genus, for example in some specimens of I. batatoides.

Choisy’s system continued in use until the 1890s when it was essentially reproduced in the account of Convolvulaceae in Die Natürlichen Pflanzenfamilien (Peter 1891). However, acceptance was never universal and Grisebach (1862b) reduced many of Choisy’s genera to sections of Ipomoea, a decision in which he was followed by Meisner (1869), Gray (1878, 1886) and others. Nevertheless, it was only in 1893 that a major generic reorganization was proposed by Hallier. The major innovation in Hallier’s (1893a, b) system was the use of pollen. He divided the Convolvulaceae into two pollen groups based on whether the pollen was smooth or spiny. The spiny pollen group, which included Ipomoea, was itself divided into two subgroups essentially on the basis of the fruit distinction proposed by Choisy for his tribes Argyreieae and Convolvuleae. The first subgroup (Echinoconieae subgroup Ipomoeeae) was composed of species with a dehiscent capsular fruit and comprised Lepistemon, Calonyction and Quamoclit as well as Ipomoea (in which Hallier included Exogonium, Pharbitis, Marcellia and Legendrea). The second subgroup (Echinoconieae subgroup Argyreieae) was characterized by its indehiscent fruit and comprised Argyreia, Rivea, Ipomoea tiliifolia and Blinkworthia. Two new genera in this second group were established: Stictocardia to accommodate I. tiliifolia and a few related species based on the prominent black leaf glands and strongly accrescent sepals and Astrochlaena (=Astripomoea Meeuse) to accommodate a group of mostly erect South African plants with stellate hairs and shortly oblong stigmas (Hallier 1893b: 159). For the first time Merremia Dennst. ex Endl. was clearly distinguished to accommodate species previously placed in Ipomoea but distinct because of their non-spiny pollen and generally white, cream or yellow flowers (Hallier 1893a). In fact, Merremia sensu Hallier represents a heterogeneous collection of species although dividing it up into natural genera is problematic (Simões et al. 2015, Simões and Staples 2017).

Hallier’s system has endured with only a few, relatively minor changes for about 125 years. A handful of new genera were established to include small, morphologically distinct splinter groups from the Ipomoeeae such as Lepistemonopsis with fleshy scales at the base of the filaments and Pentacrostigma with a 5-lobed stigma and 5-locular ovary. On the other hand the genera Calonyction, Mina and Quamoclit, all recognized by Hallier, were gradually abandoned; none was recognized by O’Donell in his various publications (O’Donell 1959a, b) although Mina is still occasionally accepted (Deroin 2001). Within the Argyreieae there has been uncertainty about the limits of various genera, notably Rivea and Turbina, the latter reincorporated in this subgroup and unique for its nearly pantropical distribution. A new genus, Paralepistemon was established to include two African species with thickened filament bases. (Lejoly and Lisowski 1986). In passing, it should be noted that a rather eccentric attempt to reclassify Convolvulaceae by Roberty (1952, 1964) has been universally rejected, like a similar earlier attempt by Rafinesque (1837, 1838a,b).

Recent phylogenetic studies point towards the acceptance of a broad concept of Ipomoea to include all Hallier’s Echinoconieae. Initial studies by Wilkin (1999) and confirmed by our own more extensive sampling (Muñoz-Rodríguez et al. 2019) have shown that even when smaller genera recognized within the tribe Ipomoeeae have strong phylogenetic support, they are nested within Ipomoea. Manos et al. (2001) showed that species with spiny pollen split into two major clades, a result confirmed by Muñoz-Rodríguez et al. (2019). One clade consists of species placed in Stictocardia, Rivea and Argyreia together with a superficially heterogeneous group of species from Ipomoea and Turbina, composed mainly (but not exclusively) of Old World species. The second clade consists of mostly (but not exclusively) New World species but includes Astripomoea, some species hitherto treated as Turbina and all Australian endemics we have sampled. Neither clade can be diagnosed by specific morphological features and it seems that there are multiple origins for many of the characters in Ipomoea including both the capsular and indehiscent fruit types as well as the different number of ovary cells, many characters thus being homoplastic. It is clear from these studies that Ipomoea as hitherto understood is not monophyletic and an expanded circumscription of Ipomoea is required to secure its monophyly (Stefanovic et al. 2003).

Our own extensive studies (Muñoz-Rodríguez et al. 2019) confirm earlier papers and support an Old World origin of Ipomoea s.l. They indicate that the recognition of a broad Ipomoea based on the presence of spiny pollen is the only logical solution that integrates monophyly and diagnosability. The alternative of dividing the whole clade into many small, formally recognized groups is not recommended due to high levels of homoplasy, lack of diagnostic characters and a complex tree model in which it is not obvious to which clades some species, not sampled for molecular data, should be placed. Consequently, we are adopting a wide concept of the genus to include Argyreia, Astripomoea, Blinkworthia, Lepistemon, Lepistemonopsis, Rivea, Stictocardia and Turbina, which are all nested within Ipomoea. However, as the genus is so large, we informally recognize certain diagnosable clades within Ipomoea to facilitate discussion and reflect the phylogenetic history of the genus. These informal clades include some traditionally recognized genera as well as newly discovered groups that contain similar looking plants and are geographically coherent. Where we have recognized informal taxa, we have been as explicit as possible about what species belong to those clades.

An inevitable result of the situation described in the previous paragraphs is that all existing infrageneric classifications of Ipomoea are to a degree unnatural and many subgroups are neither monophyletic nor well-defined, something that goes far towards explaining the instability of all previous infrageneric classifications.

Choisy (1845) divided Ipomoea into groups based on habit but, while superficially practical, this is clearly artificial and so has only been occasionally and partially adopted by subsequent authors, such as Meisner (1869) and Matuda (1964, 1965, 1966a,b). Grisebach (1862b) reincorporated Quamoclit, Calonyction and Pharbitis into Ipomoea as sections and this treatment was followed by Bentham in Bentham and Hooker (1876), although these authors also incorporated other elements such as Aniseia Choisy within Ipomoea. Clarke (1883) followed Grisebach but treated the sections as subgenera. Hallier (1893b) began the introduction of a hierarchy of infrageneric taxa by recognizing subsections as well as sections and this process was continued by House (1908b), who multiplied the number of subsections. There was then a lull in attempts at an infrageneric classification of American species (O’Donell (1941 and passim) appears to have had no interest) until a major reformulation was made by Austin (1979, 1980), who recognized an even more extensive hierarchy with sections, subsections and series. Austin’s work culminated in a detailed and very complex hierarchical classification published 16 years later (Austin and Huáman 1996).

Apart from the repeated changes of status that these infrageneric taxa have undergone resulting in many groupings being re-graded from subgenus to section to series, the increasing multiplication of infrageneric taxa illustrates the difficulties of achieving a satisfactory classification. Apart from Quamoclit, Pharbitis, Calonyction, Old World Astripomoea and eventually Batatas (Austin 1975a, 1978b), none of these groupings are entirely natural, not even Stictocardia or Arborescens (McPherson 1981), both of which comprise a diagnostic and readily identifiable core of species but with a varying number of other heterogeneous elements. The essential instability of these classifications is well illustrated by the history of series Anisomeres. This was recognized by Austin and Huáman (1996) as a distinct series but was discarded a year later in a paper dissolving the Anisomeres series (Austin 1997). We believe that any attempt to provide a subgeneric classification following a traditional Linnaean model is bound to be artificial, impractical and doomed to failure (Carine and Scotland 2002).

The history of species recognition in Ipomoea is somewhat chequered. Good taxonomic decisions always require an awareness of previous publications as well as a good understanding of the relative value of different taxonomic characters. Access to a good range of specimens and images as well as field knowledge are also useful but only a few taxonomists have been able to benefit from these. Most have worked with limited material. The new species of early authors have not always stood the test of time (Wood 2017). However, amongst those publishing a significant number of new species of neotropical Ipomoea, it is clear that both Desrousseaux and Kunth demonstrated a high degree of competence, avoiding excessive duplication of names and so most of their species have endured.

The legacy of the most important 19^th century expert on Ipomoea, the Swiss botanist Choisy, is very mixed and he was criticised even during his own lifetime. He saw a wide range of specimens in many European herbaria and was well aware of previous publications, but neither his generic nor his species concepts have lasted well. He described the same species under different names often in different genera (Choisy 1845). Ipomoea indica was described under at least eight different names, mostly in Pharbitis and I. batatas under at least four names in Batatas but also in Ipomoea. Similarly I. asarifolia was described under three names in Ipomoea, I. eriocalyx twice, once in Pharbitis once in Batatas, I. delphinioides three times in Ipomoea and so on.

The next major work was the account of Convolvulaceae prepared by Meisner (1869) for the Flora Brasiliensis and this was virtually a monograph of the family in South America. It was as unsatisfactory as Choisy’s works but for different reasons. Meisner appears to have largely discounted the names of earlier authors including those of Choisy. He superfluously redescribed species as if earlier publications did not exist - Ipomoea chrysotricha Meisn. instead of I. hirsutissima Gardner, I. obtusiloba Meisn. and I. heterotricha Meisn. instead of I. bonariensis Hook., I. riedelii Meisn. instead of I. batatoides Choisy, I. graminiformis Meisn. instead of I. schomburgkii Choisy, I. llaveana Meisn. instead of I. funis Schltd. & Cham., I. cardiosepala Meisn. instead of I. philomega (Vell.) House) and many others. The other major problem with his approach was the recognition of a large number of varieties under each species, a process that tended to obscure the boundaries of recognized species. These varieties often recognize trivial variation but, in contrast, in other cases treat distinct, often very distinct, species as mere varieties of unrelated species.

The inadequate species level taxonomy of Choisy and Meisner was not an inevitable consequence of the epoch in which they worked. Near contemporaries such as Grisebach (1862a, b, 1879) or Gray (1878, 1886 were far more reliable. However, the excessive recognition of varieties continued to bedevil species level taxonomy into the 20^th century in the publications of Kuntze (1891, 1898), Hallier (1899c) and especially of Chodat and Hassler (1905) and Hassler (1911). Hassler took this interest in infraspecific taxa to an obsessive level, providing names for every slight variant found in Paraguay. However, the publications of the North Americans, especially House (1907 and passim), Brandegee (1889 and passim) and Robinson (1891 and passim) saw a welcome decline in the number of described infraspecific taxa associated with the species they described from Mexico. However the new century did not bring a marked improvement in species delimitation. Urban (1902–3 and passim) and Standley (1924 and passim) between them described over 50 new species of Ipomoea, of which only a quarter are recognized today (Wood 2017.

From the mid-20^th century onwards, the situation has improved, partly as the result of the outstanding achievement of the Argentinian botanist Carlos O’Donell (1941, 1948a, 1948b, 1950a, 1950b, 1950c, 1952, 1953a, 1953b, 1959a, 1959b, 1959c, 1960). He described 56 species of Ipomoea of which 45 are accepted today, a success rate far higher than that of any of his predecessors (Wood 2017). Almost everything he described from Argentina is still accepted. He did much to rationalize the chaos left by Hassler, sorted out the Quamoclit Clade throughout the Americas and added significantly to our knowledge of Ipomoea in Bolivia, Peru and Brazil. The vast majority of his taxonomic decisions stand today and his judgement in an era before the internet is remarkable.

After O’Donell’s premature death in 1954, there has been a slow but steady increment of new species from the Americas. Isolated species from different countries have been published by various authors but the contributions of Dan Austin and Andy McDonald are the most significant. Accounts of Ipomoea in Panama, Ecuador, Venezuela by Austin (1976, 1982a, 1982b) have provided a framework for the study of Ipomoea in these countries. McDonald (1987a, 1995, 2001) has revised several groups of Ipomoea principally from Mexico but has also contributed with Austin and Murguía-Sánchez to the account in the Flora Mesoamerica (Austin et al. 2012). Other important contributions have been made by Carranza (1998 and passim) on Mexican Ipomoea and Liogier (1955 and passim) on those of Cuba and Hispaniola.

Inevitably the first species of Ipomoea to be recognized and catalogued from the New World were species of economic importance (I. batatas), of horticultural value (I. alba, I. indica, I. nil, I. purpurea), or were widespread conspicuous species of accessible habitats, such as I. pes-caprae, or I. violacea, which grow on seashores. All were known to pre-Linnean botanists and featured in Species Plantarum and other near contemporary works.

Geographically, the first region from where a reasonably comprehensive inventory of Ipomoea emerged was the eastern United States. Nearly all the localized species from this area had been found and described by around 1800, including species like I. pandurata, I. lacunosa and I. macrorhiza. The United States Southwest had to wait until the 1850s after the United States-Mexican war. It was only then that species from this region were discovered, principally by Wright, Lindheimer and Torrey. Most were described a quarter of a century later by Gray (1878, 1886) including I. leptophylla, I. tenuiloba, I. barbatisepala, I. lindheimeri and I. cardiophylla. After the mid 19^th century there was little new to discover in the United States, and most (but not all) described novelties were ephemeral, being shown subsequently to be conspecific with earlier species.

The Caribbean had been one of the earliest regions of botanical exploration and many of its endemic species including Ipomoea ternata, I. tenuifolia, I. repanda, I. digitata, I. clausa and I. desrousseauxii were discovered in the 18^th century, as both Jamaica and Hispaniola were visited by botanists from different European countries. Cuba was somewhat different. Although Humboldt and Bonpland visited Cuba, they did not find any of its endemic species. A few were discovered by Sagra in the 1840s, but it was only in the 1860s that the rich diversity of Ipomoea in Cuba became known after Grisebach (1862a, 1866) wrote up the many new species found by Charles Wright (the same Wright who had been active in the United States Southwest). In the years after 1870 there was a slow increment of new species from the Caribbean culminating in the collections of Eric Ekman on Cuba (I. baliocalda, I. erosa) and on Hispaniola (I. luteoviridis), although many of his supposed novelties described by Urban (1924a and passim) have proved to be synonyms of other species. The occasional new species has been found since but the inventory of Caribbean Ipomoea is now largely complete.

There were collections from Mexico in the Spanish colonial era but those of Sessé and Moçiño were not published until a hundred years later. Nevertheless, seeds sent to Spain by them and by Née were cultivated in Madrid enabling Cavanilles to describe several attractive and interesting Mexican species including Ipomoea tricolor, I. stans and I. bracteata at the end of the 18^th century. The expedition of Humboldt and Bonpland constituted the next step forward in revealing the wealth of Mexican Ipomoea. Ipomoea cholulensis, I. suffulta and I. hastigera were amongst their discoveries, as was I. arborescens, the first tree Ipomoea to be described. During the first half of the 19^th century Mairet, Andrieux, Hartweg and others added to the list of species known from Mexico, but the most important advance came with the collections of Galeotti, which greatly increased the number of known species. His discoveries were published in 1845 and included many well-known Mexican species such as I. lindenii, I. minutiflora, I. chenopodiifolia, I. pauciflora, I. suaveolens and I. proxima (Martens and Galeotti 1845).

After 1845 there was a lull in the discovery of new Mexican species for almost fifty years. However, the end of the 19^th century proved to be a golden age for botanical exploration in Mexico thanks to a series of collectors mostly from the United States, especially Palmer, Pringle, Purpus, Nelson and Brandegee, and the Mexican-Italian Casiano Conzatti. The number of recognized species doubled during this era. However, these collections did not exhaust the riches of Mexican Ipomoea and the 20^th century has seen the regular discovery of new species by collectors from both Mexico and the United States, notably H.S. Gentry, G.B. Hinton and Rogers McVaugh from the United States and the Japanese-Mexican Eizi Matuda. This trend has continued into the new century with at least eight new species described since 2000. It is too early to say whether this trend is ending but it is perhaps significant that rather few new Mexican species has been found during the course of our studies in Ipomoea.

It was mostly during the 20^th century that the Ipomoea flora of Central America was discovered and described. Although not as rich as the Mexican flora, there has been a steady increment of new species since the middle of the century including I. chiriquensis from Panama, I. magniflora from Costa Rica, I. riparum from Honduras, I. heterodoxa from Belize and I. steerei from Yucatán, mostly found by North American collectors. However, the inventory of species seems to be nearly complete, since, as with the Caribbean, nothing new has been reported since the turn of the 21^st century.

The earliest collections from South America of any importance were made by Ruiz and Pavón in Peru at the end of the 18^th century. They noted surprisingly few new species of Ipomoea but amongst them were I. ramosissima. Of far greater importance was the expedition of Humboldt and Bonpland. Having found new species in Mexico they went on to find a series of new species in the northern Andes including I. discolor and I. parasitica in Venezuela, I. capillacea in Colombia and I. abutiloides in Ecuador.

Essentially little more was discovered or described from the Andean region for well over a century apart from a few species from Venezuela (Pittier 1927, 1931), a few from Peru by Weberbauer (Ooststroom 1933), two from Bolivia (Rusby 1896, 1899) and a couple from Argentina (Grisebach 1879, Kuntze 1891).

This situation only changed after the Second World War initially as a result of O’Donell’s short career (O’Donell 1941 and passim). He significantly increased our knowledge of species in the southern Andes, describing at least six new species from Argentina (Ipomoea jujuyensis, I. rubriflora, I. lilloana, I. oranensis etc.) as novelties, two from Bolivia (I. tarijensis, I. suburceolata), two from Peru (I. velardei, I. peruviana), three from Colombia (I. colombiana, I. killipiana and I. reticulata) and one from Venezuela (I. pittieri); these all still recognised. Fieldwork by Danish botanists led to the discovery and description of three new species from Ecuador by Dan Austin. The diversity in Bolivia was, however, only revealed recently (Wood et al. 2015, 2018, Wood and Scotland 2017b), with the description of 21 new species, mostly Andean which took the total number of Ipomoea species known from that country to 109, thus putting it in third place after Brazil and Mexico for the total number of Ipomoea species recorded.

The 19^th century, in contrast, was a golden age for plant discovery in Brazil, mostly under the stimulus of the production of Martius’ Flora Brasiliensis. The roll call of collectors finding new species of Ipomoea in Brazil is composed of most famous plant collectors in Brazil in the 19^th century. They include the Germans Martius, Riedel and Sellow, the Brazilians Vellozo and Silva Manso, Blanchet and Glaziou from France, Regnell from Sweden, Gardner, Spruce and Burchell from Britain and Pohl from Austria, their achievement commemorated in species such as I. burchellii, I. regnellii, I. blanchetii, I. spruceana and I. pohlii, all still recognized Brazilian species. The result was that by about 1870 our knowledge of Ipomoea was greater in Brazil than elsewhere in South America.

After the publication of Flora Brasiliensis (Meisner 1869), there was lull in the process of discovery in Brazil, which did not really pick up again for more than a hundred years. It has only been since the 1980s that significant numbers of new species have been found and described from Brazil (Austin 1981, Simão-Bianchini and Pirani 2005 Ferreira and Miotto 2011, Vasconcelas et al. 2016, Wood et al. 2017a,d, Wood and Scotland 2017a,b). It is clear that Brazil is the richest country in South America for Ipomoea but it remains the least explored and it is the only country in the Americas from where we would expect significant numbers of new species to emerge.

As in other aspects of its history, Paraguay (and neighbouring parts of Argentina) has followed a somewhat different trajectory. Until the 1870s, the flora of this region was essentially unknown. Then came a publication by Parodi (1877) listing around 15 species of Ipomoea but in the absence of associated specimens, these names cannot mostly be linked to recognized names. Expeditions by Morong and Balansa began to reveal the diversity of Ipomoea in Paraguay, but it was a long-term Swiss resident, Emile Hassler, and Teodoro Rojas, the Paraguayan curator of his herbarium, who really discovered the Paraguayan flora and revealed the number of Ipomoea species in the country. Between them they added some 20 recognised species, all of which are endemic to the region, some extending into nearby parts of Argentina. Those species that were not recognized by Hassler himself were described subsequently by Carlos O’Donell (1948a, 1950b, 1953a), together with a number of species from Misiones and Corrientes provinces in neighbouring Argentina.

This monograph is based fundamentally on the study of herbarium specimens of Ipomoea informed by observations from morphology and molecular sequence data, fieldwork, photographs and information from literature and individual contacts throughout the Americas.

We have depended heavily on herbarium collections at Kew (K) and the Natural History Museum in London (BM), which together with material at Oxford (OXF) have formed the basis of our study. We have visited various European herbaria including Edinburgh (E), Leyden (L), Paris (P), Madrid (MA) and Stockholm (S) to view their collections of Ipomoea. During the course of visits to the United States we have seen material at (GH) and (A) at Harvard, the New York Botanical Garden (NY), the Smithsonian Institution in Washington (US), the Field Museum (F), Missouri Botanical Garden (MO) and Arizona University (ARIZ), including extensive material from Fairchild in Florida (FTG). Within Latin America, visits have been made to see herbarium collections in Cuba (HACB, HAJB), Mexico (IEB, MEXU), Ecuador (Q, QAP, QCA, QCNE, GUAY, LOJA), Peru (CIP, CUZ, USM), Bolivia (BOLV, HSB, LPB, USZ), Paraguay (FCQ, PY, SCP), Argentina (CTES, LIL) and Brazil (CEN, CPAP, HUEFS, IPA, JPB, MBM, PEUFR, R, RB, SP, UB). Help received from individuals in all these institutions is detailed in the acknowledgements at the end of this monograph. We have also received important loans of material from most of these institutions as well as from G and GOET. Photographs of herbarium specimens have also been a valuable source of information. The most important have been the images of types available through Jstor (www.jstor.org), but the websites of CRIA (splink.cria.org.br) and Reflora (reflora.jbrj.gov.br) and those associated with herbaria, including ARIZ (SEInet; swbiodiversity.org), B, BR, C, COL, E, F, MO, NY, P, PMA, US and W have all provided valuable information. We have also been sent images of important material from Geneva (G), Turin (TO), St Petersburg (LE), Vienna (W), Göttingen (GOET), Rancho Santa Ana (RSA), Montevideo (MVM) and Cambridge University (CGE). All cited acronyms are in accordance with the Index Herbariorum (http://sweetgum.nybg.org/science/ih/).

Much of the material we have been loaned has been type material or old or rare specimens and this has had important limitations on our ability to provide complete and accurate descriptions. In particular, details of the habit of many species is missing and can only be inferred. Flower colour has often been lost or modified during the drying process. It is often impossible, or at least undesirable, to dissect corollas, where only one or two are present pasted to the sheet and fragile in nature. Finally, it must be emphasized that the fruit of many species is unknown.

It is important to stress that herbarium specimens are not only a source of basic taxonomic information and an indispensable tool for species delimitation but also an essential resource for phylogenetic, ecological and other information. We have been able to use specimens for DNA sequencing, even from collections over a hundred years old, if they have been rapidly dried and retain their natural colouring. More recent, heat-dried specimens nearly always yield high-quality DNA, but there are striking exceptions, such as specimens of Ipomoea chondrosepala, which have mostly resisted repeated attempts to extract DNA. Specimen labels are another invaluable source of data. They can provide information that is not apparent from the specimen, such as flower colour, habit and size. Label information can also contain information about the general and specific habitat of the plant and can provide important facts about flowering patterns and ecology. We have used all available information of this kind to inform our descriptions and notes.

The first author has had many years of fieldwork during which he has collected Ipomoea. However, it is only since about 2008 that he has made careful efforts to collect, photograph and study the genus. Most of his fieldwork in South America has been carried out in Bolivia but important visits have been made to Argentina with the help of Hector Keller, to Brazil with the help of Luciano de Queiroz and to Paraguay with the help of Rosa Degen. This fieldwork has been very important in enhancing our understanding of the variation in species and in providing details of their habit and habitat. A consequence is that Bolivia is the only country from where we have near complete molecular sampling, a near complete collection of photographs of living plants and a good understanding of the phenology of different Ipomoea species. It is fortunate that there are 109 recorded species in Bolivia making it the third most species-rich country for Ipomoea in the Americas after Mexico and Brazil.

We have also benefitted from observations and in particular images sent to us by individuals over the years. We are particularly grateful to Maira Tatiana Martinez, Alfredo Fuentes, Alexander Parada, Julia Gutiérrez, Modesto Zarate and Daniel Soto (Bolivia), Moises Mendoza and Hibert Huaylla (Bolivia and Brazil), Hector Keller and Keith Ferguson (Argentina), Gilberto Morillo (Venezuela), Regis E. Bastian, Teresa Buril and Ray Harley (Brazil), Mario Giogetta (Bolivia and Argentina), Erin Tripp (Mexico), Jhon Infante Betancour (Colombia), Rémi Girault (French Guiana), Ramona Oviedo and José Luis Gómez (Cuba). We have also benefitted from images of living plants shown on a number of websites, especially Tropicos (tropicos.org) and SEInet.

We have made full use of a wide range of literature as cited in the list of references. This includes regional, national and local floras and checklists (WCSP (1917), for example) as well as taxonomic works. We have consulted field guides and similar works when we have become aware of their existence. They often provide specific habitat and field identification information not readily available elsewhere. We have also made occasional use of information on the internet, but only if it seems reliable. We have scanned literature for examples of illustrations of species to supplement those prepared specifically for this project.

Nowhere in biology is the disparity between theory and practice more evident than at the level of species. In an influential and widely cited contribution Kevin de Queiroz (2005, 2007) proposed the ‘unified species concept’ to treat existence of species as ‘separately evolving metapopulation lineages’ as the only necessary property of species and that the plethora of species concepts in existence merely represent different lines of evidence relevant to assessing lineage separation. In this way Queiroz (2005, 2007) separated the theoretical idea of what species are from the operational criteria of how to discover them. An important issue for the recognition of species is that as lineages diverge they can become distinguishable as separate species with diagnostic characters of fixed traits. Species can evolve distinctive ecologies and they can pass through polyphyletic, paraphyletic, and monophyletic stages in terms of their component genes. The problem is that these changes do not all necessarily occur, or if they do occur, do not do so at the same time and they do not even necessarily occur in a regular order (Queiroz et al. 1998). What this dynamic system of divergence means is that there is a certain pragmatic and heuristic nature to species delimitation whereby, although the expectation is that many species are clearly monophyletic, there will be other situations where the estimate of the degree of separation comes down to taxonomic judgement. Therefore, in this monograph we consider species to be separately evolving metapopulation lineages and in the discussion that follows we describe the operational criteria we used to infer, delimit and make taxonomic judgements about species boundaries. From experience gained studying Ipomoea we consider that species delimitation is usually relatively straightforward, given a representative sample of specimens and an understanding of the important diagnostic characters in the genus. Thus, a first task is to gain access to a wide range of specimens for comparison purposes, followed by a study of publications by reliable taxonomists who have worked on the genus. At the same time as specimens are studied morphologically, representative leaf samples from each putative species are sequenced for a few DNA barcode markers to provide an independent data source to corroborate or refute a species hypothesis based on morphological analysis or literature sources. Species delimitation is facilitated if DNA sequencing and specimen examination take place nearly simultaneously but this is only possible when DNA samples can be extracted from the available material. Conceptually as discussed above, we follow Queiroz (2005, 2007), whose framework includes the idea that disagreements about the limits of species are especially prevalent in those species at the active interface of speciation. This explains why many species have universally agreed boundaries whereas others are more difficult to interpret. We take the view that, in the context of taxonomy, for those instances where species delimitation is particularly problematic, it is best to flag up the variation, make a pragmatic, discursive and explicit taxonomic decision and move onto those other species delimitation problems that taxonomy can readily solve. This is especially true in large tropical groups such as Ipomoea in which many of the taxonomic problems can be readily solved by having a good sample of specimens combined with good knowledge of the group’s literature and nomenclature. Our view is that species and species delimitation can be viewed as a heuristic allowing an approach to problem solving or discovery that employs a practical method not guaranteed to be optimal or perfect, but sufficient for the immediate goals. To a greater extent than in Convolvulus (Wood et al. 2015), Evolvulus L. or even Jacquemontia, most species of Ipomoea are, as a generalization, well-defined. Hybridization is rarely reported, and that principally in the Batatas Clade. Claims by Austin that I. leucantha and I. grandifolia are of hybrid origin require corroboration. There are, of course, species complexes where delimitation is difficult, as intermediates occur between recognized species and these may be hybrids but a lot of the difficulty faced by the taxonomist arises from the lack of available material for study. Almost 20 species in the following account are only known from the type collection, another 50 or so are only known from less than five collections, and not all of these have been available for study. In about 20 cases, we have not seen authentic material and have relied on images and original descriptions to describe and delimit species. Inevitably, we are tentative in our decisions on the validity of some individual species. Examples include I. leucantha mentioned above as well as several species in Clade A1 (Figure 1; Muñoz-Rodríguez et al. 2019), such as I. vivianae or I. pseudomalvaeoides. A particular case is the pantropical I. indica. Molecular studies show that this is polyphyletic (Muñoz-Rodríguez et al. 2019, supplementary data 3), but it is clear that a more exhaustive study with extensive sampling is necessary before the components of this entity can be unravelled.

We have tried to make use of so-called conservative characters in accepting species and the value of these is discussed in the notes that follow. Pollination syndromes as reflected in corolla shape and to some extent in colouring and the structure of the androecium are seen as important for species delimitation (Rosas-Guerrero et al. 2011). Ideally each species will be delimited by a combination of distinctive characters. Species defined by a single character, unless it is exceptionally strong, are regarded with suspicion. We do take into account ecological and geographical factors. We would expect different species to have different ecological requirements or occupy a different geographical area from obviously related species. Where morphology, ecology and geography intergrade we have indicated by notes and in some cases by the recognition of subspecies. We have used insights from molecular sequence data, keeping separate apparently similar species like I. marginisepala and I. cardiophylla or I. asarifolia and I. paludicola where uniting them would render them non-monophyletic. We have been reassured in our species level decisions by molecular data in the recognition of numerous species, for example, I. huayllae, I. graniticola, I. chiquitensis and I. juliagutierreziae from Bolivia and I. kraholandica and I. chapadensis from Brazil. Molecular data has been less helpful for species delimitation within some of the relatively large clades such as Clades A, 1–2 (Figure 1) where multiple accessions of several species are not resolved partly due to the limited variation provided by regions used.

The concept of a subspecies is retained for taxa which are morphologically distinct throughout most of their range but whose characters overlap in regions where the ranges of the two taxa meet. We accept that some species recognized in the following account could have been treated as subspecies of a closely related species, but in many cases, the number of specimens seen is so few that it would be premature to make this decision. Subspecific status is, therefore, reserved pragmatically for taxa of which we have seen many examples. Subspecies are keyed out when there are three more recognized subspecies for a particular species.

Apart from subspecies, other infraspecific categories are not formally recognized in this account with the exception of two varieties. Most varieties, formas and subformas recognized by previous authors have little value and often do little more than recognize minor variations of corolla colour, indumentum or leaf shape. Some varieties, however, have long been recognized and, where these are historically significant or readily recognized, we have drawn attention to them in the notes that follow the species descriptions and made comments about their distinctive characters and distribution. We accept that some readers may wish to continue recognizing and using these varietal names. We understand varieties as morphologically distinct populations that occur sporadically over part or all of the range of a species. Although varieties may be restricted to a specific area they do not occupy a distinct geographical region with populations overlapping with those from another distinct geographical area. Sporadic occurrence is an important criterion in the recognition of varietal, rather than subspecific status.

Results from molecular sequencing and phylogenetic analysis have been of great value in our research at many levels (Wood et al. 2015; 2017a, b, d; 2018) and have enabled us to delimit Ipomoea as a genus, facilitating the study of its evolution (Muñoz-Rodríguez et al. 2018, 2019). It has enabled us to recognize major radiations in South America within Clade A1 (Figure 1) and in the Caribbean within Clade A2. It has confirmed the monophyly of some groups previously recognized on morphological grounds, such as Calonyction, Quamoclit, Astripomoea and Batatas (Muñoz-Rodríguez et al. 2019, supplementary data 3–8). Other accepted groupings, such as Argyreia, Pharbitis, Stictocardia and the Arborescens group are shown only to be monophyletic if certain species are excluded. Conversely, it has demonstrated that some recognized groups are not monophyletic (Turbina for example) and that Rivea is nested within Argyreia. Importantly the phylogenetic framework we have developed provides a context in which to interpret and understand the evolution of the many species of Ipomoea that lay outside the previously recognized segregate genera.

DNA sequencing and phylogenetic analysis has been valuable at the species level too. It has confirmed the monophyly of many species and has also drawn attention to the existence of unrecognized new species. We have many examples of this, such as the “discovery” of Ipomoea kraholandica in Brazil or I. lactifera in Bolivia, this last especially interesting as DNA confirmed it as belonging to the Batatas Clade and sister to the Old World species, Ipomoea littoralis. Sequence data has shown some species thought to be distinct are conspecific with others from different geographical areas, for example, I. acanthocarpa from Africa is the same species as I. piurensis from America, while I. lindenii from mainland America is the same as the Jamaican endemic I. cyanantha. In both these examples multiple specimens of the supposedly distinct species form a largely unresolved single clade confirming our morphological observations that the species are indistinguishable. DNA has also demonstrated that some species pairs thought to be possibly conspecific are indeed different; I. paludicola is distinct from I. asarifolia, I. marginisepala from I. cardiophylla and I. pterocaulis from I. jalapa. In these examples specimens from the two species do not form a clade separate from all other species. DNA has also shown that Ipomoea indica is not monophyletic and so consists of more than one entity, although we have not yet been able to unravel this complex. It has highlighted misidentifications when wrongly named specimens appear in parts of the tree separate from the clade where they belong. It has also provided a phylogenetic context to enable the interpretation of incomplete specimens, which lack diagnostic morphological information. However, molecular sequencing using ITS has severe limitations which are well documented, not least lack of resolution and support. For this reason, we have always used our ITS phylogeny in conjunction with hypotheses based on morphological characters. Nevertheless, we have been reassured that all major clades identified in our ITS tree are also inferred from the analysis of single-copy nuclear regions and of whole chloroplast genomes (Muñoz-Rodríguez et al. 2019).

Figure 1 summarises the phylogeny of Ipomoea and shows the genus is divided into two clades of similar size. These are labelled for ease of reference but are not formally recognized. The two clades are dominated by species from the Old World and New World respectively but with many exceptions. The Old World Clade (OWC) consists of species previously placed in Argyreia, Rivea, Stictocardia and Lepistemon as well as many always included in Ipomoea. The New World Clade (NWC) consists of an early diverging grade of Old World elements and a species-rich clade dominated by species from the Americas, but which also includes all species endemic to Australia. The existence of this fundamental split within Ipomoea had been posited by previous research (Miller et al. 1999 Manos et al. 2001) although based on much poorer taxon sampling.

Both NWC and OWC contain elements that were recognized previously as genera and appear as smaller clades within NWC or OWC. The only previously recognized genus that is represented by native species in both NWC and OWC is Turbina, although most of its species are in OWC. Turbina is polyphyletic containing several heterogeneous elements and is consequently rejected. Similarly, we reject the New World genus Exogonium as it was founded on a hypocrateriform corolla adapted for bird pollination and this character is homoplastic occurring in various different clades within NWC. In NWC, species formerly grouped under the names Arborescens, Batatas, Pharbitis, Calonyction and Quamoclit all form small clades which more or less coincide with their traditional circumscription and so are used by us as names for the corresponding clades. In OWC the generic names Argyreia, Astripomoea, Stictocardia and Lepistemon form clades of varying sizes and we continue to use these names for these distinct clades. Unlike the New World clades recognized above, these Old World clades vary considerably in size, Argyreia having around 125 species (including Rivea), Stictocardia around ten and Lepistemon only two so barely meriting recognition. All these nine clades which are assigned traditional names are more or less diagnosable using combinations of morphological characters.

NWC comprises about 450 species. Apart from a few species previously placed in Turbina, all species belong to Echinoconieae subgroup Ipomoeeae in Hallier’s (1893a, b) classification. Most (All?) Australian endemics belong to this clade and it is also well represented in Africa. Within NWC, two very large clades are recognizable. One clade (mostly South American but including Batatas) roughly coincides with subgenus Eriospermum (Hallier f.) Verdc. ex D.F. Austin as defined by Austin and Huáman (1996), although Austin included many elements which do not belong, such as Ipomoea rubens, I. lindenii, I. violacea, I. imperati, I. magnifolia, I. habeliana etc. (Muñoz-Rodríguez et al. 2019). We refer to this as Clade A. There is a second very large, mainly Mexican, clade that has not previously been recognized which includes Austin’s subgenera Ipomoea and Quamoclit as well as some other species. We refer to this as Clade B. Clades C, D, E and F represent smaller clades within NWC, this last essentially African with one New World endemic (I. habeliana).

Our studies have revealed many smaller clades to which a traditional name cannot be readily attached. The two largest are both in Clade A of NWC and we refer to these as Clades A1 and A2. Some of the species in Clade A1 were treated as series Jalapa by Austin and Huáman, but in a very inconsistent way. It is found throughout the neotropics but is most diverse in South America. The Arborescens group form a small clade within A1. Clade A2 is also found throughout the neotropics but is particularly important in the Caribbean, as nearly all the 25 endemic species of that region belong to it. Elements of this clade were referred to as Microsticta by McPherson (1980) and as series Eriospermum by Austin and Huáman (1996). Both Clade A1 and Clade A2 are usually recognizable morphologically, the former by its pubescent corolla and rather soft, flattish sepals and the latter by its usually glabrous corolla and coriaceous, often convex, ovate to elliptic sepals. We note that there are a few exceptions in both these clades and that several of these diagnostic characters are homoplastic in other parts of Ipomoea. Clade A3 is a small clade comprising the Batatas group. Ipomoea cryptica is sister to this clade in the nuclear phylogeny but not in the chloroplast phylogeny.

Apart from Pharbitis, Calonyction and Quamoclit, there are several small clades which are more or less diagnosable morphologically within Clade B. There is a small clade (Species 328–334) of seven species centred on Ipomoea costellata assigned the name Pedatisecta by House (1908b) characterized by digitately divided leaves. These were treated as part of Sect. Leptocallis by McDonald (1995) but the name Leptocallis has to refer to a quite different small clade (Species 280–288) centred on I. capillacea, perhaps characterized by tuberous storage roots. The most distinct small clade in Clade B consists of five species characterized by pinnatifid leaves and centred on I. stans (Species 275–279). These were included in Sect Tyrianthinae by McDonald (2001) but this name cannot be used for this clade as the type, I. orizabensis, belongs to a different clade. Since the six small clades discussed here account for only a small proportion of the species in Clade B, we have avoided any formal recognition of these names.

Clade C also contains a number of small clades which are more or less diagnosable morphologically or geographically, although the best known species, Ipomoea pes-caprae, belongs to a clade of Australian species. A small clade of four species (Species 345–348) centred on I. asarifolia can be recognized by their very unequal, transversely muricate sepals. Another small clade of South American species consisting of perhaps eight species centred on I. maurandioides (Species 356–363) that can be recognized by their glabrous indumentum, unequal sepals and often trailing habit.

The Old World Clade (OWC) contains around 350 species mostly from the palaeotropics. It includes most species treated as Echinoconieae subgroup Argyreieae by Hallier including all species placed in Argyreia, Rivea (which is nested within Argyreia), Stictocardia and some species placed in Turbina. Only a few relatively small clades are composed of neotropical species. Much the largest is the clade of around 12 species centred on I. corymbosa but with morphologically very disparate elements, including I. ochracea, I. regnellii, I. crinicalyx, I. cuscoenesis and I. daturiflora.

There are important practical implications from our molecular results. Since there is no obvious or close correlation between morphological characters and the Ipomoea phylogeny, it is currently impossible to propose an infrageneric classification along traditional lines. Although most clades cannot be defined morphologically, they do have certain morphological tendencies, which we have highlighted and discussed in the notes that precede the description of the species in each clade. As noted above, some of the smaller clades are well-defined and, where this is the case, their distinctive morphological features are indicated. We have also tentatively used molecular results to inform the placement of individual species within clades.

It should be stressed that we have faced a problem that we share with previous botanists working on the classification of Ipomoea. Some species are not available for study or sequencing and so cannot be assigned unequivocally to a clade. In this situation, we have inferred the position of species from their morphology. Most placements will be uncontroversial but in a few cases they are little more than guesses. The notes following each species indicate where placement is particularly uncertain.

Ipomoea is a tropical genus and this is reflected in its distribution in the Americas with few species found north or south of 30 degrees latitude. The main exception lies in the Eastern United States where several species, I. coccinea, I. lacunosa, I. sagittata and I. pandurata, extend north to at least 35 degrees, I. repanda as far as 43°N in Ontario, Canada. The complete absence of Ipomoea from California in the west apart from a few introduced ornamentals (as well as in central Chile) suggests that it cannot tolerate a Mediterranean climate with arid summers and cool wet winters.

Within the neotropics Ipomoea is widely distributed but is noticeably less diverse in the equatorial region with relatively few species in Amazonia, Ecuador (Austin 1982a) or Colombia (Bernal et al. 2015). Although a partial explanation lies in the low diversity of Ipomoea species generally in rain forest, it does not account for the lack of species diversity in the dry forests of the Caribbean coasts or of the inter-Andean valleys such as the Colombian Magdalena. Species diversity rises as one moves away from the Equator and the countries with the greatest diversity of species lie mostly within the 15 to 30 degrees of latitude, notably Mexico and Brazil, both large countries with extensive subtropical dry forest. Some smaller countries in these latitudes, such as Paraguay (Wood et al. 2017c), Bolivia (Wood et al. 2015) or Cuba (Wood and Scotland 2017c) are proportionately as rich.

Most species of Ipomoea are relatively localized in their distribution often being found in a single region or country. However, there is a large set of species (I. alba, I. batatas, I. cairica, I. carnea subsp. fistulosa, I. corymbosa, I. hederifolia, I. indica, I. muricata, I. nil, I. purpurea, I. quamoclit and I. tricolor) that occur around cultivation or in disturbed places near settlements throughout the tropics and are found in almost every country of the Americas with a tropical climate. To this group should be added some other pantropical species that are also widespread but absent from many countries including I. acanthocarpa, I. aquatica, I. asarifolia, I. fimbriosepala, I. mauritiana, I. setifera and I. triloba. All these pantropical species occur sporadically, occasionally abundantly, in different neotropical countries but there is little geographical patterning to their distribution. A similar pattern can be observed in the palaeotropics. Of the 26 species recorded for the Flora of the Mascarenes (Bosser and Heine 2000 all but one also occur in the Americas. Equally all but two species recorded from Hawaii are also present on the American continent.

Of species never found in the Old World, Ipomoea aristolochiifolia is probably the most widespread, being found from Argentina north to Mexico, although it is absent from the Caribbean islands. Other very widespread species include I. philomega, I. batatoides, I. ramosissima and I. regnellii but, apart from I. ramosissima, none extends into Argentina and all peter out as they enter Mexico. Two species, I. dumetorum and I. clavata extend along the Andean Chain from Argentina or Bolivia north to Mexico but are absent elsewhere. More frequent are species that extend from the United States or Mexico southwards to northern South America. These include I. capillacea, I. cholulensis and I. lindenii that are restricted to the mountain chains and I. minutiflora, I. meyeri, I. trifida and I. tiliacea which are common in the Caribbean (except I. minutiflora) and Central America extending into northern South America, in the case of I. tiliacea south along the eastern edge of Brazil almost to Uruguay. Of some interest are two upland species, I. plummerae and I. pubescens, common around the 20–30° latitude in both hemispheres but largely absent from intermediate equatorial regions.

Ipomoea plummerae and I. pubescens are not the only species with disjunct distributions. Ipomoea crinicalyx and I. amnicola are also amphitropical in distribution but there is suspicion that the latter has been introduced into the northern hemisphere. Several annual species like I. parasitica, I. heptaphylla, I. longeramosa and I. neurocephala are very scattered in their distribution, being known from many countries but, with the exception of I. longeramosa in NE Brazil, from only one or few collections in each case. The occurrence of the South American I. subrevoluta on the Isla de Juventud (Pinos) in Cuba and also on Trinidad is remarkable but it perhaps arrived as a result of the movement of migratory water birds. Ipomoea thurberi also has a curious distribution with isolated populations in Guatemala and Nicaragua which are disjunct from each other as well as from the main population in northern Mexico and Arizona. In South America remarkable disjunctions are noted for species found on isolated granite domes around the Amazon. Ipomoea chiquitensis and I. graniticola are known from a few locations separated by many thousands of kilometres (Wood et al. 2017c). The apparent disjunctions in the distribution of two species found in the Amazon basin, I. amazonica and I. velutinifolia can be explained by inadequate collecting in areas separating known locations. The most inexplicable disjunction, however, is that of I. eremnobrocha known from the Cerro Campana in Panama and from a number of locations in NE Brazil.

Throughout the Americas many species are endemic to single countries with a good number of species endemic to single localities or to a very restricted area. Clearly the two largest countries, Brazil and Mexico, each with about 60 endemic species, have the greatest numbers of single country endemics. Scattered endemic species are found in most Andean countries with much the greatest numbers in Bolivia (c. 20) but the arbitrary nature of political boundaries tends to reduce the gross figures for individual countries. There are few species endemic to the small Central American republics although four are endemic to the Panama-Guatemala region. The large Caribbean islands are also major centres of endemism. We recognize 17 species as endemic to Cuba, seven to Hispaniola and four to Jamaica. Additionally there are a number of near endemics on these islands. In contrast, species endemic to small islands or island groups are few and we recognize only four, Ipomoea sphenophylla on St Eustatius, I. steudelii on Puerto Rico, I. tuboides on Hawaii and I. habelana on the Galapagos, the last two on several islands in their respective archipelagos and, perhaps coincidentally, both adapted for moth pollination.

It is harder to discern concentrations of endemic species in particular regions of the large continental countries, particularly in Mexico, where endemic species occur in scattered locations over much of the country. However, there is evidence that the greatest concentrations of endemics are in the seasonally arid regions of South West Mexico (McDonald 1991), with a lesser centre in the central northern plateau. Much the same is true for South America but the Chapada de Veadeiros (Brazil) is home to at least four endemic species and the Sierra de Amambay (Paraguay) to at least three. Both these locations are also home to several other very rare species which extend only to a few nearby locations. Another very rich area comprises the lower eastern slopes of the Andes near the border of Argentina and Bolivia. This is exceptionally diverse in terms of local endemic species with at least nine species endemic to the area.

It is equally difficult to discern clear examples of endemism in particular biomes except for some extreme examples such as seashores. Clearly there are many species endemic to Seasonally Dry Forest and to Cerrado but as the former includes many distinct variants and the latter very different physiognomies from campo limpo to cerradão, the notion of endemism is not very easy to apply except in a very loose sense. Specific examples of habitat preferences are indicated after species descriptions, where these are reliably known.

Precise information about the ecology of many species is unavailable so it is difficult to provide anything approaching a comprehensive account of the habitat requirements of many neotropical species. Certainly, Ipomoea species grow in many different habitats and it is clear that most habitats host species specific to that habitat.

The most typical beach species are Ipomoea pes-caprae, I. imperati and I. littoralis (in Hawaii) but others occur on coastal sands including I. tiliifolia and some forms of I. batatas. There is some evidence that the fruits of some of these species can survive for long periods in salt water (Miryeganeh et al. 2014) and it has been suggested that in the case of I. pes-caprae the persistent pedicel actually aids seed dispersal. The world distribution of these species and that of I. violacea, which often grows in mangrove swamp, strongly suggests that their dispersal is mediated through ocean currents. This may be the explanation of how the salt marsh species, I. sagittata made it to Europe in prehistoric times. Ocean currents may also partially explain the distribution of Ipomoea indica and I. triloba as both show a predilection for islands, although the former is also readily spread by broken shoots as a result of trampling by cattle. There is also an interesting group composed of species that are not strictly maritime but are often found in the proximity of the coast, although all occur, sometimes abundantly inland; these include Ipomoea tiliacea, I. mauritiana, I. digitata, I. asarifolia, I. macrorhiza and I. jalapa.

Some species are characteristic of freshwater habitats and are often specialized in their requirements. The only true aquatic is the introduced Ipomoea aquatica, which roots on mud and sometimes has extensive floating stems. Ipomoea subrevoluta usually grows by small streams in grassy plain whereas I. rubens is more typical of the borders of larger rivers or small lakes. Ipomoea paludicola, I. schomburgkii and I. pittieri favour flooded pampa whereas I. paludosa is characteristic of swampy hollows in the cerrados. Ipomoea fimbriosepala, I. setifera and I. neei are often found near water. The widespread species I. alba appears to favour disturbed scrubby gullies which are permanently or seasonally moist, when it grows as an apparently native species. The natural distribution of I. carnea subsp. fistulosa is obscured by its presence as an escape from cultivation but it appears native in swamp in the Parana basin of South America and perhaps elsewhere.

The lack of diversity of Ipomoea in rain forest does not mean that there are no characteristic species in this habitat. The best indicator of rainforest in the genus is I. philomega, which is found in evergreen forest at low altitudes throughout the Americas. Other typical species that are more local in their distribution include I. amazonica, I. velutinfolia, I. santillanii, I. splendor-sylvae whereas I. aurantiaca, I. chondrosepala, I. regnellii, I. squamosa, I. batatoides and I. reticulata also occur in rainforest but are not restricted to this habitat. The near absence of several otherwise widespread species from the Amazon basin is also interesting. Ipomoea hederifolia and I. carnea subsp. fistulosa are almost completely absent from Amazonia.

Cloud forest is another wet forest habitat where Ipomoea is relatively poorly represented. Cloud forest occurs from slightly below 1000 m to at least 2500 m along the Andes from Bolivia northwards, in the Brazilian Atlantic forest and in Central America. Probably the most widespread cloud forest species is I. lindenii, which grows from Bolivia to southern Mexico with an outlying station in Jamaica. Other cloud forest species are much more local but include I. austrobrasiliensis from the Brazilian Atlantic Forest, I. magnifolia, I. inaccessa and I. odontophylla from the Bolivian Andes, I. retropilosa from Colombia and Venezuela, I. chiriquensis, I. isthmica from Panama and Costa Rica and I. chenopodiifolia from Guatemala and Mexico. Other species may occur in coffee plantations, which are often created from areas of former cloud forest including the widespread I. aristolochiifolia.

High altitude species are even rarer and very few species occur above about 2500 m. The only species that might occur in paramo is Ipomoea capillacea while, in puna or at least subpuna, the only species recorded are I. plummerae and I. pubescens. Both have a disjunct amphitropical distribution occurring in Mexico and the United States Southwest as well as South America. Ipomoea plummerae reaches 4000 m in Bolivia.

Ipomoea species are tolerant of drought and several are recorded from desert. In South America I. incarnata is the best adapted to arid conditions occurring in the coastal deserts of Peru and the Colombian Guajira as well as the Caatinga of NE Brazil. Other indicators of very arid conditions in South America are I. nationis from Peru, I. verruculosa from Venezuela and I. sericosepala from Brazil and Bolivia. In North America, Felger et al. (2012) record some 13 species from the Sonora desert region of Mexico-Arizona, listing I. cardiophylla, I. costellata, I. cristulata and I. ternifolia as typical of this habitat. Most tree species from the Arborescens clade in both South and North America favour arid habitats but are more typical of dry deciduous forest than true desert.

Of some interest are morphological adaptations found in several species growing in dry habitats. One such occurs in the coastal lomas of Peru and the northern Atacama of Chile. Here forms of Ipomoea dumetorum, I. nil and I. purpurea occur with short, erect stems, very unlike the normal long twining stems found in other habitats. The Galapagos Islands comprise another arid habitat where there occur extreme forms of I. muricata and I. incarnata, once treated as distinct species under the names respectively of I. tubiflora and I. linearifolia. In the former the fleshy teeth of the stems are largely suppressed while the latter presents with very narrow leaves. In the Sonora Desert in Mexico, forms of I. cristuluta occur with erect, woody virgate stems, a facies very different from the normal herbaceous, twining stems. Perhaps the most remarkable is the dwarf form of the usually lowland I. platensis which grows in arid situations at over 2000 m in the Argentinian Andes. (Figure 83).

Desert merges into dry grassland, particularly in North America. Erect and, less commonly, trailing species of Ipomoea are characteristic of grassland habitats. There are relatively few examples from North America, I. leptophylla being the only widespread prairie species but several other North American species are clearly adapted to the grassland habitat, including I. longifolia and the Mexican endemic I. durangensis. However, it is in the South American cerrados that a great number of grassland species have evolved. Erect species occur in different clades and include I. hirsutissima, I. malvaeoides and I. cuneifolia and several others from Clade A1, I. argentea and I. paulistana from Clade A2 and I. squamisepala and I. pinifolia from Clade C. Trailing species are also common including I. descolei, I. psammophila and I. langsdorfii, I. burchellii, I. goyazensis and I. procumbens.

Thorn scrub merging into seasonally dry forest is another important semi-arid habitat, which is common throughout much of tropical America. Ipomoea is at its most diverse in this habitat. In South America the relatively widespread species I. amnicola, I. megapotamica, I. incarnata and I. abutiloides are good indicators of this habitat. However, each of these dry forest regions has its own set of localized species, I. argentinica, I. oranensis and I. schulziana where the chaco meets the Andes, I. brasiliana, I. longibracteolata, I. marcellia and others in NE Brazil. Ipomoea verruculosa in the dry coastal woodland of Venezuela, I. pauciflora and I. velardei in Ecuador and Peru. Dry forest species are also noted from the Caribbean Islands, I. carolina from Cuba, for example, but it is in Mexico and Central America that very large numbers are recorded as growing in dry forest, usually pine or oak woodland, either wholly deciduous or partially so. All the tree species (from both North and South America), lianas like I. bombycina and numerous other species are recorded from this habitat. The roll call of dry forest species from Mexico is long and includes such relatively common species as I. orizabensis, I. pedicellaris, I. praecana, I. seducta, I. lobata and many others.

Ipomoea species tend to avoid closed forest but occur along streams, by tracks and roads and often favour rock outcrops where the forest cover is broken. Species diversity is greatest in deciduous forest, possibly because there is more plentiful light during the dry season (McDonald 1991). This could be an explanation for why some dry forest species flower in the dry season at a time when they are leafless. This is a particular feature of the tree species in general, some Mexican species such as I. tehuantepecensis, I. pseudoracemosa, I. concolor and I. pruinosa, but of relatively few South American species wth the exception of I. juliagutierreziae and I. schulziana.

Rocks provide a specialized habitat for some species. In Mexico, cliffs or “crags” are often cited as the habitat for Ipomoea rupicola, I. chilopsidis, I. teotitlanica, I. seeania and I. concolor whereas in South America the only species cited from a similar habitat is I. killipiana. The geological composition of the cliffs is not usually recorded but volcanic rocks are mentioned for I. seeania and limestone for I. teotitlanica. Limestone, however, is often cited for plants from the Caribbean including I. montecristina, I. praecox and I. fuchsioides from Cuba, the last two characteristic of limestone towers locally known as mogotes. It is also cited for several species from Hispaniola including I. digitata and I. desrousseauxii. Ipomoea luteoviridis is recorded from serpentine outcrops in Hispaniola but we are unaware of any other American species with this habitat preference. A few species are noted from lava flows, notably I. tuboides from Hawaii, but several Mexican species are recorded on pedregales including I. orizabensis and I. dumetorum. In South America the most commonly recorded specialized rock habitat consists of granite domes and platforms, which outcrop sporadically in dry forest and cerrados on the pre-Cambrian shield. The commonest species of this habitat are I. bonariensis and I. maurandioides, but neither is restricted to granite. More restricted geographically and geologically and often very disjunct in their distribution are I. caloneura, I. chiquitensis and I. graniticola, the last being found in isolated locations in Bolivia, Brazil and Paraguay. Ipomoea leprieurii is locally frequent on granite outcrops in French Guiana and neighbouring parts of Brazil while I. marabaensis, I. scopulina and I. fasciculata are currently known only as pin-point endemics.

Ipomoea species are also frequent in secondary scrub and in disturbed places around settlements. This is the kind of habitat where the widespread pantropical species are often found. Ipomoea indica, I. nil, I. hederifolia, I. purpurea and I. cairica are rarely found far away from human habitation and I. alba, I. cairica, I. tricolor, I. indica, I. quamoclit and I. carnea subsp. fistulosa are sometimes clearly garden escapes. The same is true for many species of the Batatas clade. Ipomoea tiliacea, I. triloba, I. cordatotriloba, I. australis, I. leucantha, I. grandifolia and I. trifida are all recorded as characteristic of disturbed bushy ground and are rare in truly natural habitats.

Many species have a distinct, relatively short flowering season. The only country where details are documented, albeit superficially is Bolivia (Wood et al. 2015). Similar details are largely unknown from other countries although information about 12 Mexican species is provided by Chemás-Jaramillo and Bullock (2005). The short flowering season is at least a partial explanation for why some species are rarely collected and so are only known from one or two examples.

Certain generalisations, however, are possible. The erect cerrado and grassland species with a stout xylopodium often come into flower soon after the start of the spring rains, possibly being stimulated into growth and flowering by the fire that often precedes the onset of rain. Annual species, in contrast, use the moist summer season for growth and come into flower towards the end of the summer, their flowers often persisting long into the winter dry season (see Chemás-Jaramillo and Bullock (2005) for examples from Mexico). Most dry forest and semi-desert species flower during the summer rainy season, taking advantage of the short wet period to produce their flowers. One subset, however, prefers to flower in the height of the dry season when they are leafless so their seeds are mature when the rains eventually begin (Ipomoea schulziana, I. juliagutierreziae). Plants of flooded pampas flower after the waters recede during the winter. There is no clear pattern amongst species of moist forest. The archetypical rain forest species, I. philomega flowers at the height of the summer but other moist forest species such as I. regnellii and I. cryptica prefer the winter.

There are many individual subtleties, which need careful observation and recording before any explanation can be provided. In Eastern Bolivia in areas of a similar altitude and climate, the first author has observed the following sequence, although these observations may be partially dependent on the date of the onset of rain. To see flowering specimens of I. hirsutissima, I. cerradoensis and I. psammophila, it is best to visit in October and November; to find I. schomburgkii, I. aprica, I. caloneura and I. paulistana it is best to look in December or January; to find I. graniticola and I. densibracteata February to early March would be best; March to early April would be good for I. amnicola, I. abutiloides and I. megapotamica; April to June would be good to find I. bonariensis, I. argentinica, I. rubens, I. bahiensis and I. cordatotriloba; to find I. ramosissima, I. setifera, I. paludicola or I. eriocalyx June or July would be best, while July or August might be best for I. regnellii, I. lactifera and I. cryptica. Finally you should note that you might find I. maurandioides in flower at almost any season.

Ipomoea species are commonly named “Morning Glory” because the flowers of several cultivated species, notably I. indica, open at dawn and close before midday. However, while this observation may be a useful generalization, it is only a partial truth. Much depends on the strength of the sun and many morning-flowering species will continue in flower well into the afternoon on a dull day. Conversely night-flowering species, such as I. alba, I. muricata and I. violacea may remain open during clouded, sunless days. These observations indicate that research suggesting different species flower for a specific number of hours (Chemás-Jaramillo and Bullock 2005) should be treated with caution. However, there is no doubt about the truth of their observation that the flowers of some species, especially robust perennials, such as I. ampullacea, I. bracteata and I. pedicellaris, remain open for much longer periods than those of more slender species.

Much the most important species of Ipomoea economically is I. batatas, the sweet potato, which is reported to be amongst the ten most important staple food crops worldwide (Woolf 1992, FAO 2017). Although clearly of American origin it is widely cultivated in almost all tropical and subtropical countries for its root tubers (storage roots). The largest contemporary producer is China but much of Chinese production is used as animal fodder (FAO 2017). It has a number of important advantages as a human food. It is second only to the potato in productivity per hectare. It is more drought resistant than many important staple crops such as maize. The common orange-fleshed varieties are an outstanding source of Vitamin A and have significant quantities of Beta-carotene, potassium and various other elements important for human nutrition (Kurabachew 2015). Indeed per gram it is richer in potassium than bananas (USDA 2017. The purple-fleshed varieties have enjoyed a recent vogue as brain food but it is unclear whether this is merely a fashion fad or based on sound evidence.

Other species of Ipomoea produce root tubers but there are only occasional reports of their use, usually as a famine food. Amongst species whose tubers are reported to be used for food are I. leptophylla, I. pubescens, I. pandurata (Haddock et al. 2015), I. plummerae (Gutiérrez-R 2016) and I. serrana (Vasconcelas et al. 2016).

The leaves of some species of Ipomoea are used as a vegetable. Much the most important is I. aquatica, the water spinach or kangkong, which is widely used as a stir-fry vegetable in South East Asia, although it has not achieved much popularity outside the region. The leaves of other species are occasionally used as vegetables, including I. batatas itself and apparently I. littoralis (Austin 1991b), although it is unclear whether they enjoy general use or are a resort at times of famine. It is possible that the leaves of other species could be used as a vegetable but the leaves of some species are potentially harmful (Meira et al. 2012). Ipomoea malvaeoides and I. carnea subsp. fistulosa, for example, are avoided even by goats and are unpalatable, if not actually poisonous, to animals and presumably to humans.

Various species of Ipomoea are cultivated as garden ornamentals. In extra-tropical countries, relatively quick growing annual species are favoured, particularly I. indica, I. purpurea, I. nil, I. quamoclit and I. tricolor. In tropical countries, perennials are more common. The most conspicuous is I. carnea subsp. fistulosa, which is widely cultivated for its erect habit and profuse flowers. Ipomoea cairica is often planted to cover walls and unattractive bushes. Ipomoea alba and I. muricata are also sometimes grown in gardens and on boundary fences. Ipomoea horsfalliae is a widely planted liana that is grown in many tropical countries for its attractive red flowers, but is not reported to set seed and so is never naturalized. Ipomoea quamoclit and, less commonly, I. lobata are also grown quite frequently and sometimes become naturalised. There are occasional reports of the cultivation of other species including I. nervosa, I. pauciflora and I. intrapilosa but this is not common practice.

Various species of Ipomoea have had medicinal uses since pre-Colombian times, broadly for two purposes. The seeds of several species are known for their hallucinogenic properties as they contain small quantities of LSD-like substances (Steiner and Leistner 2018). Amongst the species used as a hallucinogen are I. tricolor “Heavenly Blue”, I. purpurea, I. alba, I. corymbosa and I. nervosa. The roots of several species have been used as a purgative and marketed under the name “jalapa”. Ipomoea purga is the best-known species used for this purpose but others such as I. simulans, I. orizabensis and I. jalapa are sometimes reported as having similar properties, although their medical value requires confirmation. Meira et al. (2012) document many actual and potential medical uses of Ipomoea species.

Species of Ipomoea may be annual or perennial, herbaceous or woody, twining (or at least scrambling), erect, decumbent or prostrate. All of these characters are potentially useful in species delimitation and are used in the keys. It is useful, for example, to distinguish between lianas and scandent herbs or between prostrate or erect herbs but the distinctions need to be treated with caution. Many species have a woody rootstock and herbaceous stems, which may or may not be woody at the base. Stems may become somewhat woody with age. Twining plants may be trailing in the absence of shrubs to climb on. We have also avoided the use of the term vine as it is sometimes used to mean a woody climber (like the grape vine), so almost a synonym of liana, and sometimes to mean a relatively slender twining plant.

Annual species are characterized by having fibrous roots and typically flower in the late rainy season (tropical summer) as they require sufficient time to reach maturity after the onset of rains. In the herbarium, in the absence of roots, annuals can often be identified by their slender habit and the presence of mature capsules on flowering specimens. Perennial species, in contrast, are relatively stout and often lack mature capsules on flowering specimens or are almost entirely without corollas on fruiting specimens. It is possible that some normally annual species perenniate under suitable circumstances, especially in areas with no distinct dry season. There are no known erect annual species. Annual species are not found in Clades A1 or A2. In contrast they are well-represented in the Batatas (A3 in part), Pharbitis (B1 in part), Quamoclit (B2 in part) and the Pedatisecta Clades (B2 in part).

The majority of species are twining perennial herbs or lianas with petiolate, ovate, cordate leaves. The inflorescence is formed of pedunculate axillary cymes, the cymose structure usually being very obvious, although the cymes are sometimes reduced to single flowers. There is a tendency for some of the lianas to develop inflorescences on short leafy branchlets, rather than from the axils of the stem leaves.

Somewhat similar is a less well-defined assembly of essentially trailing plants. At one extreme these species root at the nodes and form extensive mats, in one case (Ipomoea aquatica) extending its stems to float on shallow water. Two widespread submaritime species, I. pes-caprae and I. imperati, are good examples of this growth form. More common are trailing species that do not root at the nodes. They usually grow in open, often sandy inland habitats. These trailing species often have shortly petiolate, elliptic leaves rounded to truncate at the base combined with axillary cymose inflorescences, these sometimes being shortly pedunculate. These trailing plants are, thus, apparently intermediate morphologically between the true climbers and the erect species. Some trailing species are morphologically indistinguishable from the climbers, the prostrate habit apparently the consequence of the absence of suitable plants to climb. Ipomoea maurandioides, a South American species principally of rock outcrops, is one such example.

The erect habit is usually associated with subsessile, oblong, lanceolate, or oblong-elliptic cuneate-based leaves with a terminal inflorescence, the upper leaves clearly bract-like and the pedicels and peduncles reduced so the inflorescence is subracemose or even subspicate in form. Species with this habit occur mostly in open grasslands and especially in the cerrados of South America. Most species produce annual stems from a tough woody perennial subterranean xylopodium, which is resistant to fire, a characteristic and perhaps defining feature of these habitats. Erect species are found in many different clades but are unknown in the Batatas, Quamoclit and Pharbitis Clades and rare in Clade B.

The erect habit is also associated with a number of shrubs and small trees often treated as Section Arborescens. These usually (always?) have white latex and often flower when leafless or nearly leafless. The inflorescence often develops on short branchlets and is not obviously axillary and cymose in structure. The corolla is white with a dark centre, subcampanulate to funnel-form in shape and possibly bat-pollinated (Felger and Austin 2005). Species with these characteristics mostly occur in very dry forest along the mountain chains of Mexico, Central America and the Andes and are completely absent from Brazil and the Caribbean.

Much the most widespread and common erect species, Ipomoea carnea subsp. fistulosa fits none of the above characteristics, having ovate cordate leaves and pink flowers in axillary cymes but its uniqueness is perhaps a consequence of its close relationship with Ipomoea carnea subsp. carnea which is a characteristic climbing species, from which it is presumably diverged.

Although annual species are generally known to have fibrous roots, little reliable information is available about most of the perennial species. Erect species of the cerrado nearly always arise from a woody xylopodium but this is known to vary considerably in form and development from species to species. Ipomoea hirsutissima, for example, has very large somewhat woody tuberous roots. Similar storage roots are seen in other species in Clade A1 including I. lilloana (Figure 15D) and I. opulifolia (Figure 15E). The best-known species for its tuberous rootstock is, of course, the edible I. batatas, but storage roots occur in many different clades throughout the genus, such as I. bonariensis and I. platensis in Clade A2, this last sometimes cultivated as a succulent. Those of I. pubescens and I. plummerae in Clade B are sometimes eaten, while those of I. pandurata and I. leptophylla in Clade C are noted for their size. Other species have tubers which can be used medicinally, notably I. purga and I. jalapa. However, for the vast majority of species there is no accurate information about their rootstock. Although this character may prove to be of economic importance in the future and is significant in discussions around the origin of the sweet potato (Muñoz-Rodríguez et al. 2019), it can be little used at the present time in species delimitation.

White latex is recorded as present in many species and is sometimes abundant, notably in trees and lianas, including species in the Arborescens and Calonyction Clades as well as in the aptly named Ipomoea lactifera. However, its presence often goes unrecorded and it may be more or less obvious according to climatic conditions.

Stems may be entirely herbaceous, woody in the lower parts and herbaceous above, or entirely woody except for the new growth. Stems may be glabrous or variously hirsute, the indumentum usually being similar to that of the peduncles, petioles and leaves, especially the abaxial surface of the leaves. There is a tendency for older stems to be somewhat glabrescent. Unusual features of the stem include distinct wings (Ipomoea pterocaulis, I. splendor-sylvae, I. subalata, I. kahloae), squamose dark glands (I. balioclada), warty protuberances (I. verruculosa, I. tuboides), spinules (I. spinulifera), soft spines (I. setosa), soft fleshy teeth (I. muricata, I. alba, I. parasitica) and granulose protuberances (I. granulosa).

Species may be glabrous or variously hirsute. There is a good deal of intra-species variation and this has often proved to be an unsatisfactory character in species delimitation. Many species or varieties have been recognized over the years based on the presence or absence of hairs and have subsequently been abandoned. Despite this important proviso, many species have a characteristic indumentum which is readily recognized. Species which are always glabrous in their vegetative parts form a long list, as do those which are characteristically sericeous or tomentose. A sericeous indumentum is characteristic of almost all species previously placed in Argyreia, Rivea, Turbina and Stictocardia as well as many that have always been included in Ipomoea. Some unusual indumentum types include:

• Stellate hairs. These are characteristic of certain species notably Ipomoea bonariensis from South America, I. scopulorum from Mexico and I. luteoviridis from Hispaniola. In cases where they are mixed with simple hairs they may be very difficult to observe and pass unnoticed. They are also characteristic of the Astripomoea Clade, which is restricted to Africa.

• T-shaped hairs. Ipomoea malpighipila was named on the basis of the presence of T-shaped hairs. They are not reported from other species, except the related I. aemilii, and are difficult to observe even in these species.

• Scattered long fine hairs. Ipomoea clavata, I. dolichopoda.

• Density and appearance. Many species are densely hairy especially on young stems and the abaxial surface of leaves but sometimes on all vegetative parts. Where hairs are dense the leaves are often white or grey in colour and characteristic of the species. This kind of indumentum is not always easy to define and is sometimes described as canescent, sericeous, tomentellous, tomentose or densely pubescent by different authors.

• Gland dots. Distinct gland dots are found in some species, especially on the abaxial leaf surface but sometimes on other vegetative parts or even the corolla. They usually appear as dark dots and are so characteristic of I. tiliifolia that they are often regarded as a defining characterstic of the Stictocardia Clade (Austin and Demissew 1997). They occur sporadically elsewhere as in some specimens of I. megapotamica, I. reticulata and I. batatoides. As white dots they are characteristic of I. eremnobrocha and the related species I. isthmica and I. peteri.

These have been reported in many species including Ipomoea alba, I. batatas, I. bonariensis, I. carnea, I. indica, I. leptophylla, I. mauritiana, I. muricata, I. pes-caprae and I. tuboides (Keeler 1977, 1980, 1985, Keeler and Kaul 1979, Mondal et al. 2013, Meeuse and Welman 2000). These are usually found on the petioles, at the base of the leaf where it joins the petiole or on the sepals. It is postulated that they attract ants which help to protect the plant from predators. However, they are not readily observed and their taxonomic value is uncertain as they are not necessarily constant in a particular species (for an example, see the discussion about I. indica in Keeler (1985). The case of I. tuboides is particularly interesting as there are no native ants in Hawaii, suggesting perhaps that the nectaries evolved in the ancestor of this species before it was dispersed to Hawaii from the American mainland.

Leaves are exstipulate but a few species have pseudo-stipules (notably Ipomoea cairica, I. fissifolia and I. quamoclit), formed by modified leaves or prophylls. Leaf size can be distinctive but difficult to quantify diagnostically. Large leaves are a feature of a few species such as Ipomoea ampullacea, I. magnifolia and I. philomega whereas small leaves are characteristic of many annual species but also of some perennials such as I. hartwegii and I. rupicola.

Leaf shape is mostly related to habit with almost all climbing species having ovate to deltoid leaves with a truncate, cordate or sagittate base. Elliptic leaves are rare and mostly found in trailing species. Lanceolate, oblong or oblong-elliptic leaves are mostly a feature of erect species. Some unusual shapes occur, such as the strap-shaped leaves of I. tenuissima.

Leaves may be entire or variously divided. Pinnate leaves are only present in Ipomoea quamoclit, and pinnatifid to lyrate-dentate leaves in a few Mexican species (I. ancisa, I. sescossiana, I. tacambarensis, I. stans). A much larger number of species have leaves palmately lobed. The number of lobes, usually 3 or 5, occasionally more, and the depth of lobing are often characteristic of a particular species. However, leaf lobing is often an inconstant character, many species having entire-leaved forms or forms that intergrade with the normally lobed forms. The leaves of some, such as I. bonariensis, I. clausa, I. microdactyla or I. mauritiana are notoriously variable in form. A relatively small number of species have leaves palmately divided into separate leaflets and this character is usually constant. Species which present forms with both lobed leaves and leaves divided into separate leaflets occur in only a very few species (I. cairica, I. bonariensis, I. homotrichoidea).

The leaf base is sometimes distinctive, particularly in those species that have leaves with strongly cordate or strongly cuneate bases. Sagittate or hastate leaves are also often distinct but may intergrade with the more common cordate leaf base. Rounded leaf bases often intergrade with shallowly cordate or truncate leaf bases and are difficult to characterize.

The leaf margins are usually entire to slightly undulate but a few species have distinctly dentate leaves (I. odontophylla, I. schaffneri, I. noctuliflora, I. ignava, I. peruviana, I. descolei and I. erosa). A few species may have 1–several rather large teeth on the margins, usually towards the base (I. acanthocarpa, I. dumetorum, I. eriocalyx). In the majority of species the leaf apex is acute to acuminate, although the actual tip may be somewhat obtuse. The tips are commonly mucronate but in a few cases the midrib extends as a mucro several millimetres in length (I. walteri). In a few species the apex is distinctly retuse (I. pes-caprae).

In general, petiole length is of little significance except that short or absent petioles correlate with an erect habit and elongate leaf shape as noted earlier. One curious feature is the fusion of the petiole and the peduncle at least for part of their length (I. connata, I. bracteata, I. dumosa).

Most inflorescences consist of cymes that arise from the leaf axils. Cymes are nearly always solitary but are very variable in the number of flowers. In many species the cymes are reduced to a single flower while in others the cymes may be compounded with up to 15 or more flowers. The number of flowers in the cyme is often a useful although somewhat imprecise taxonomic character.

Not all inflorescences are obviously cymose in structure, some are more or less corymbose (especially in the Quamoclit Clade) or racemose (e.g. Ipomoea bombycina, I. reticulata, I. corymbosa) or umbellate (some forms of I. batatas), even appearing paniculate in some forms of I. lineolata or I. philomega. In quite a few species, the pedicels are very short so the inflorescence is subcapitate in form. In the Arborescens Clade and also in a number of woody lianas, the inflorescence arises on short leafy (bracteate) branchlets with no obvious cymose structure.

We have generally avoided using the term bract since in most twining or trailing species, the bracts are not clearly differentiated from the leaves, the cymes arising in the axils of the leaves which function as bracts. In the erect species and also in some or the arborescent species where the inflorescence is either terminal or borne on small branchlets bracts are more clearly differentiated from leaves, typically smaller and narrower and diminishing in size towards the branch tips and, in this situation, we have used the term bract. Some authors, however, use the term bract for the very different structures that arise at the inflorescence branching points or at the base of the pedicel in unbranched inflorescences. We refer to these as bracteoles, only rarely differentiating between primary bracteoles (at the first branching point) or secondary bracteoles (at the higher branching points) as these rarely differ in any significant way. In many species the bracteoles are inconspicuous and caducous (and have never been observed in a few species), but in others they are prominent and persistent, especially in the Pharbitis Clade, and occasionally even forming an involucre around the flowers where the pedicels are very short, notably in I. neurocephala and I. involucrata.

In the majority of species the bracteoles are small (< 3 mm long), often linear, lanceolate or scale-like and caducous. In a few species, Ipomoea blanchetii is an example, we have not observed bracteoles in any specimen available to us. In others, they are relatively persistent, particularly in species, with a subcapitate inflorescence. These include I. indica, I. villifera, I. mairetii, I. argentinica, I. asplundii, I. chrysocalyx, I. racemosa, I. amazonica, I. eriocalyx, I. setifera, I. fimbriosepala, I. burchellii, I. pohlii and I. mcvaughii. In a very few species the bracteoles are expanded, persistent and form an involucre around the inflorescence as in I. neurocephala, I. involucrata, I. bracteata and I. suffulta.

Peduncles may be short or long and the length is sometimes significant. Most species with a terminal inflorescence have very short peduncles and pedicels. However, some trailing or twining species are also remarkable for their relatively short peduncles. These include Ipomoea eriocalyx and a miscellaneous group of other species, such as I. lindenii, I. chapadensis, I. riparum and I. chrysocalyx but is most common in Clade A2. Species in this clade with very short peduncles include I. microdonta, I. lachnea and I. calophylla from the Caribbean, I. goyazensis from South America, I. isthmica and I. heterodoxa from Central America and I. pseudoracemosa, I. pruinosa, I. conzattii and I. tehuantepecensis from Mexico. Many of these species with short peduncles also have short pedicels so the whole axillary inflorescence is very compact. However, there is also a group of species with relatively long peduncles but a subcapitate inflorescence in which the flowers are borne on short pedicels. This is particularly characteristic of the Pharbitis Clade (I. indica, I. neurocephala, I. mairetii, I. lambii and I. villifera) but is also noteworthy amongst many unrelated species including I. racemosa, I. amazonica, I. argentinica, I. bahiensis, I. eriocalyx, I. fasciculata, I. exserta and I. batatas. Species with pedunculate subcapitate inflorescences often but not always have a bracteolate inflorescence. Very long peduncles are also distinctive in species such as I. marcellia, I. macdonaldii, I. longibarbis, and I. austrobrasiliensis. Unusually long pedicels are rarely apparent but are a feature of I. pedicellaris and its allies which include I. regnellii, I. lindenii and I. tentaculifera, these inflorescences appearing very lax. Unusual features of the peduncle include the winged peduncles of I. decemcornuta and I. kahloae, the peduncle fused with the petiole for some of its length (I. connata, I. dumosa, I. bracteata) and the peduncle that passes through the leaf sinus (I. aristolochiifolia, I. huayllae). Very occasionally pedicels are unusually slender and coiled (I. heptaphylla, I. tenera).

The calyx is formed of five overlapping, free sepals. The two outer sepals are usually similar in size and form as are the two inner sepals, which often have relatively broad, scarious, glabrous margins. The middle sepal is intermediate in size and shape and is commonly asymmetrically scarious. The sepals are often of considerable taxonomic significance and constitute important conserved characters at the species level. The differences in size and shape between the inner and outer sepals are often of great significance. The apex is frequently especially diagnostic. Many species have mucronate sepals, but the mucros are often caducous so some or even all sepals may appear muticous or retuse. Also important is the abaxial surface of the outer sepals which may show all kinds of variation in indumentum, venation and surface which can be smooth, muricate or armed with soft spines. In a few species notably in the Arborescens Clade, the presence of hairs on the adaxial surface is significant. As observed by Hallier (1893a), the sepals are accrescent in fruit, more especially so in the lianas such as I. brasiliana or I. tiliifolia, sometimes doubling their size after anthesis and becoming wider so sepals which were lanceolate at anthesis may become ovate in fruit. They may also enclose or nearly enclose the capsule.

Many sepals display unusual features including:

• Very unequal sepals: I. anisomeres, I. cryptica, I. squamosa, I. asarifolia, I. paludicola, I. maurandioides, I. macedoi.

• Adaxial (inner) surface hirsute: Arborescens Clade, I. longibracteolata, I. magna.

• Subterminal awns: all species in the Quamoclit Clade.

• Sepals terminating in a long awn: I. alba, I. muricata, I. nil, I. hederacea. Sepals of some other species, such as I. incarnata, may be interpreted as terminating in an awn.

• Sepals with fleshy spine-like trichomes: I. crinicalyx, I. echinocalyx, I. altoamazonica, I. silvicola, I. setosa, I. tentaculifera, I. lozanii (smaller than in other species),

• Sepals with a prominent abaxial appendage, I. rosea, I. bahiensis; I. decemcornuta.

• Sepals with swollen abaxial tumour: I. appendiculata.

• Sepals with 1–2 prominent black abaxial glands: I. hieronymi, I. megapotamica.

• Sepals muricate: I. plummeae, I. capillacea, I. madrensis, I. aristolochiifolia, I. pedicellaris, I. obscura, I. ochracea, I. cairica, I. asarifolia, I. paludicola, I. procurrens, I. coriacea.

• Sepals with prominent longitudinal ribs: I. fimbriosepala, I. setifera, I. parvibracteolata, I panduata.

• Sepals with fimbriate margins: I. tenera, I. sidifolia (sometimes).

• Sepals with a prominent cordate base: I. macedoi, I. apodiensis, I. pantanalensis, I. pubescens, I. lindheimeri.

The great diversity of sepal form is curious and not easily explained. It has been suggested that the development of coriaceous and large sepals may have evolved in response to the need to protect nectar glands from robber insects. (McDonald 1991).

The corolla is most commonly funnel-shaped, but is quite often campanulate, or hypocrateriform, or sometimes suburceolate, the limb usually prominent, entire or shallowly lobed but occasionally deeply lobed, or much reduced and present only as five indistinct teeth. The corolla exterior has five prominent midpetaline bands, which may be more darkly coloured and/or more pubescent than other parts of the corolla exterior. The corolla is very variable in size from less than 1 cm long in species like I. eriocarpa or I. minutiflora to over 10 cm in length in species like I. jalapa, I. megalantha, I. parvibracteolata, I. subalata and I. pterocaulis. Size is an unsatisfactory character at one level because of its variability within individual species, but is nonetheless often characteristic of a particular species.

Corolla shape is usually, perhaps always, related to pollination. The commonest corolla shape consists of a very short subcylindrical basal tube which is then gradually widened to the mouth. Corollas of this type are described as funnel-shaped, are usually, pink, sometimes blue or white, in colour and are apparently pollinated by bees. The limb is entire, undulate or shallowly (very rarely deeply) lobed. When the corolla is very short, the tube is more abruptly widened from the base and is campanulate in form. This is characteristic of some species in the Batatas Clade and also of small-flowered species with a cream corolla, such as Ipomoea reticulata, I. corymbosa and I. syringiifolia. This kind of corolla tends to intergrade with the common funnel-shaped corolla. The corolla of the Arborescens Clade and some other, mostly woody liana species is shortly funnel-shaped (almost campanulate), white or white with a dark purple centre. These flowers may be bat-pollinated (McDonald 1991: 73, Felger and Austin 2005, Queiroz et al. 2015) but confirmation is needed in most cases.

Other corolla shapes are less common. A hypocrateriform or salver-shaped corolla in which the nearly cylindrical corolla tube is only slightly widened at the mouth is associated with red flowers, exserted stamens and bird pollination. This corolla type is characteristic of the Quamoclit Clade but is also fairly common in the Clade A2 in South America (Ipomoea exserta, I. longistaminea, I. ana-mariae, I. verruculosa), and especially the Caribbean (I. argentifolia, I. digitata, I. microdactyla, I. steudelii). In Mexico and northern South America it is more commonly associated with Clade B in the Pharbitis Clade (I. jamaicensis) and elsewhere (I. bracteata, I. dumosa, I. chenopodiifolia, I. retropilosa, I. tubulata). Occasionally the corolla limb is very deeply lobed as in I. repanda, I. hastigera, I. electrina (which is orange, rather than red). An occasional variation is the suburceolate corolla, in which the corolla tube is essentially cylindrical but somewhat swollen in the middle and with a short corolla limb consisting of small teeth. Ipomoea suburceolata from Bolivia, I. lobata and I. tehuantepecensis from Mexico and I. praecox from Cuba have flowers of this kind. nother variation is found in plants with a white or pale blue corolla in which the tube is exceptionally long. This type of corolla is associated with night-flowering hawk moth pollinated species. The best-known species of this type is I. alba but there are various others with similar corollas including I. habeliana, I. violacea, I. tuboides, I. scopulorum, I. riparum, I. santillanii, I. chiriquensis, I. ampullacea, I. macdonaldii and I. lottiae. Species with this kind of corolla are notably more common on oceanic islands and in Mesoamerica and Mexico than elsewhere.

Corolla colour. Field and herbarium observations of flower colour need to be treated with caution. Flowers change colour during the course of the day, most obviously in the case of Ipomoea nil, which is blue when fresh but turns pink as it ages and appears pink in herbarium specimens. Equally, one collector’s purple is another collector’s pink or lilac or even red. Although the great majority of species have a corolla colour that is generally described as pink, there are many exceptions. White flowers (often with a dark centre) are characteristic of the Arborescens Clade and of several other woody liana species, such as I. magna, I. longibracteolata, I. brasiliana and I. paradae, and are in some cases pollinated by bats. Night-flowering moth pollinated species typically with a hypocrateriform corolla, such as I. alba, I. santillanii, I. habeliana, I. violacea, I. ampullacea have pure white corollas. Campanulate or funnel-shaped white flowers are noted for many different species in different clades but are more common in the Batatas Clade (I. lactifera, I. lacunosa), Clade A1 (I. cerradoensis, I. macrorhiza, I. langsdorfii, I. vivianae, for example) and Clade A2 (I. proxima, I. suaveolens, I. pruinosa) but occasionally occur elsewhere (I. imperati). Many usually pink-flowered species are recorded as sometimes being white-flowered (I. acanthocarpa, I. bahiensis, I. carnea). Slightly different are those species with creamy or violet-tinged flowers such as I. lindenii, I. corymbosa, I. saopaulista, I. minutiflora and I. syringiifolia. Truly yellow flowers are rare in American Ipomoea but include I. ochracea, I. longeramosa and I. lutea. There are many subtle variations between red and pink. Red flowers being principally a feature of the Quamoclit Clade, some Caribbean species (I. montecristina, I. microdactyla, I. repanda and a few South American species notably I. cavalcantei). Some corollas are described as purple and include forms of I. indica, I. cuzcoensis and I. magnifolia. Blue flowers also occur and are often associated with a white corolla tube. I. hederacea, I. nil, I. aristolochiifolia, I. tricolor, I. marginisepala and I. cardiophylla are species with this corolla colour.

Corolla indumentum. The indumentum of the corolla exterior is best observed on buds as there is some evidence that hairs are caducous in some species as the corolla matures. Hairs are often difficult to see on open corollas but are best searched for at the tips of the midpetaline bands. Although previous studies have not seen corolla indumentum as particularly important taxonomically, we have found it of great significance both at species and clade level. It is nearly always constant in a particular species, exceptions being very rare and their existence raising doubts about the circumscription of the species in the few cases where it has been noted (Ipomoea lindenii, I. wolcottiana, I. brasiliana). All species of the Quamoclit and Batatas Clades have corollas glabrous on the exterior. All species in Clade A2 have coriaceous sepals and glabrous corollas (except I. discolor). All species in the very large Jalapa radiation (Species 1–83) have pubescent corollas.

The stamens are of little taxonomic value. They are always five and may be included or exserted. If they are included they are unequal with two noticeably longer than the other three but, if exserted or near exserted, they are subequal in length. The filaments are slightly expanded near the base but are occasionally thickened and subtriangular as in Lepistemon and some forms of Ipomoea batatoides. The filaments are always glandular pilose at the base. In a few species hairs are reported to extend upwards along the filament and this has been used as a diagnostic character in the Batatas Clade. (Austin 1978b).

The pollen of Ipomoea is always globose and pantoporate with large supratectal elements that form acute or blunt spines. The presence of these echinulate supratectal elements is the diagnostic synapomorphy for Ipomoea within Convolvulaceae. Within this general pollen-type subtle variations are visible in the size of the pollen grains, in the number and shape of pores, in the number and structure of the supratectal elements, in the structure of the area surrounding the pores, the presence or absence of ‘basal cushions’ sensu Wilkin (1993) at the base of the supratectal elements and the extent of columellae in different parts of the pollen (Sengupta 1972, Pedraza 1983, Wilkin, 1993). In addition, individual pollen grains may look very different depending on whether the opercula or aperture membranes remain intact (Figure 10A, C–H) or not (Figure 10B) after acetolysis.

Our own survey of Ipomoea pollen confirms these previous studies demonstrating continuous variation in pollen morphology with little, if any, discrete variation that correlates with phylogeny. The attempt by Wilkin (1993) to correlate results with a broader infrageneric classification of Ipomoea was made in the pre-molecular era and does not correspond closely with our molecular results. Nevertheless, although pollen in itself is of little phylogenetic or taxonomic value within Ipomoea, a few broad generalisations can be cautiously made. The pollen of species in Clade A (Figure 9, A–C) usually has fewer supratectal elements (spines) than the pollen of species in other clades. Pollen in Clades B and C (Figures 9D, E, 10C–H) often shows a regular pattern of 4–6 supratectal elements per pore as exemplified by Figure 10C and 10G but this is not always the case (Figure 10A, H). The pollen of I. alba (Figure 10D) and related species in the Calonyction Clade (species 271–274) have characteristic stout, rounded gemmiform spines rather than the usual acute spines, but similar blunt spines are also found in other more distantly related species such as I. dumosa (Figure 10E).

In summary, the pollen of Ipomoea is characterised by echinulate supratectal elements, showing a number of features that vary continuously and some specific morphologies that are homoplastic.

The style is elongate, equalling or extended slightly beyond the anthers and nearly always glabrous, even in species with a hirsute ovary. The only exception we are aware of is Ipomoea sidifolia, in which the hairs extend for a short distance upwards from the ovary. The style is usually included in the corolla but is exserted in species with a hypocrateriform corolla. The stigmas are characteristically biglobose, that is they are bilobed with each lobe globose and appearing fused. They sometimes appear simply globose. Triglobose stigmas are characteristic of the Pharbitis Clade but are not reported from all species in the clade. Somewhat elongate stigmas are reported from African species placed in Astripomoea Clade but also occur in three species of the Arborescens Clade: I. pauciflora, I. populina and I. wolcottiana.

The ovary is narrowly ovoid in shape and usually glabrous. A pubescent or comose ovary is rare and only commonly found in the Batatas Clade. Most species have a bilocular ovary with two ovules in each chamber. This correlates with a biglobose stigma. A few species (Pharbitis Clade) have a trilocular ovary each chamber with two ovules, this correlating with a trilobed stigma. In species of the Quamoclit Clade, in Rivea, Stictocardia and most species placed in Argyreia, the ovary is 4-locular but with a single ovule in each chamber. Very rarely other arrangements are noted. In Ipomoea decasperma (and I. longituba Hallier f. from Madagascar) the ovary is 5-locular with two ovules per chamber but it is not clear whether this is constant in all examples of these species. Ipomoea gilana is reported to have a trilocular ovary.

The fruit may be an indehiscent, woody or somewhat fleshy structure or formed by a dehiscent capsule. In species with an indehiscent fruit, this is usually globose to ellipsoid in shape and may contain up to four seeds except in those species placed in Turbina where 1–2 seeds only are present. Indehiscent fruits are glabrous but some species placed in Argyreia have mealy fruits. In those species with a capsular fruit, the capsules may be globose, ovoid or conical in shape. Capsules are usually muticous but species with a prominent rostrate apex formed by the persistent style base are common. Most capsules are completely glabrous but in a few species, they are pubescent, pilose or comose, this correlating with a hirsute ovary (Ipomoea velutinifolia, I. dubia, I. sidifolia, I. dasycarpa, many annual species of the Batatas Clade). In the majority of species the capsule is bilocular with up to four seeds, though often less as a result of abortion. There are several exceptions. In the Pharbitis Clade capsules are usually trilocular and 6-seeded. Very rarely capsules have up to 10 seeds (I. decasperma). In the Quamoclit Clade the capsules are 4-locular but with only four seeds.

Seeds (Figure 11) are typically broadly oblong in outline and vary in size from species to species. Their colour (when ripe) can vary from black to varying shapes of brown, sometimes being distinctly reddish-brown. They can be completely glabrous, minutely covered in very short hairs (tomentellous), only visible under a microscope, pubescent, tomentose or, in many species, with prominent, usually white hairs which develop on the angles of the seeds, In a few cases the seeds are completely covered in matted woolly hairs (Ipomoea bombycina, I. eremnobrocha, I. isthmica, I. macrorhiza, I. jalapa). Although important in diagnosing species and species groups, the value of seeds as a taxonomic character is somewhat diminished by a number of factors. The seeds of many species are unknown; in some the marginal hairs are caducous so may appear absent (I. psammophila) and in others there may be more variability than can be demonstrated from the few fruiting specimens known (I. jalapa).

Key A1: Species with soft fleshy spines on the sepals and/or peduncles

Key A2: Species with erect stems

Key A3: Species with leaves divided digitately to, or near the base, into five or more lobes or segments

Key A4: Species with very long sepals, mostly exceeding 2 cm in length

Key A5: Species with coriaceous, convex, usually glabrous sepals

Key A6: Species with a subcylindrical corolla tube and (usually) exserted stamens

Key A7: Species with small flowers, the corolla < 3 cm long

Key A8: Plants with a glabrous white corolla > 3 cm long (check buds).

Key A9: Plants with subcapitate inflorescences

Key A10: Trailing, climbing or twining plants with a pubescent corolla > 3.5 cm long

Key A1

Species with soft fleshy spines on the sepals and/or peduncles (Figure 15B). Excluded are species where soft spines are only on the stem, such as Ipomoea muricata and I. parasitica as these teeth occur occasionally in other species such as I. alba.

Key A2

Erect species. Perennial herbs or subshrubs growing in open habitats. Leaves subsessile (petioles usually < 1 cm), linear, lanceolate, ovate or oblong in shape, base attenuate or cuneate, rarely rounded, never cordate. Sepals various. Inflorescence usually terminal on the stem, often subspicate or subracemose in form but occasionally branched and arising from the upper leaf axils. Corolla shape and colour varied but never hypocrateriform (except I. cavalcantei) or suburceolate. Capsule and seeds varied.

Key A3

Digitate-leaved species with leaves divided to or near the base into 5 or more segments. Excluded are species with all or most leaves 3-lobed or divided to halfway or less.

Key A4

Species with very long sepals, mostly exceeding 2 cm in length

Key A5

Species with coriaceous sepals. Perennial erect, trailing or twining herbs or woody lianas, erect, trailing or twining, stellate hairs sometimes present. Leaves lobed or entire. Sepals coriaceous, convex, subequal, usually glabrous but sometimes indumentum from pedicels extends onto lower half of outer sepals. Corolla glabrous (except I. discolor), funnel-shaped with included stamens or hypocrateriform or suburceolate with exserted stamens. Capsule 4-seeded, seeds commonly with prominent, long marginal hairs.

Key A6

Species with a subcylindrical corolla tube and exserted stamens. Perennial or annual herbs of varying habit and leaf shape. Corolla subcylindrical, the tube scarcely widened upwards; stamens exserted.

Key A7

Species with small flowers, the corolla < 3 cm long

Key A8

Plants with a white or yellow glabrous corolla. Included are tree-like shrubs or lianas with white flowers and dark purplish or pinkish centres, which have not been keyed out earlier.

If sepals with fleshy spines go to Key A1.

If corolla with cylindrical tube and exserted stamens: Go to Key A6.

If corolla < 3 cm long go to Key A7.

If an erect plant with sessile/subsessile leaves go to Key A2.

If plant with coriaceous, convex sepals, go to Key A5.

Key A9

Plants with flowers in subcapitate inflorescences. Inflorescence pedunculate but flowers on reduced pedicels so clustered in a head-like inflorescence, the bracteoles often persistent.

Key A10

Trailing, climbing or twining plants not in Keys A1–9 with corolla > 3.5 cm long, pubescent on the exterior. Buds should be checked carefully as pubescence is more obvious at this stage. On mature flowers check near the apex of the midpetaline bands.

Key A11

Trailing and twining plants not in Keys A1–9 with a glabrous corolla, > 3.5 cm long.