The Bulletin of the British Museum ( Natural History ), instituted in 1949, is issued in four scientific series, Botany, Entomology, Geology (incorporating Mineralogy) and Zoology, and an Historical series.

The Botany Series is edited in the Museum's Department of Botany Keeper of Botany: Mr J. F. M. Cannon

Editor of Bulletin: Mr J. R. Laundon

Assistant Editor: Dr A. J. Harrington

Assistant Editor: Mrs M.J. West

Papers in the Bulletin are primarily the results of research carried out on the unique and ever- growing collections of the Museum, both by the scientific staff and by specialists from elsewhere who make use of the Museum's resources. Many of the papers are works of reference that will remain indispensable for years to come.

A volume contains about 256 pages, made up of two numbers: published Spring and Autumn. Subscriptions may be placed for one or more of the series on an annual basis. Individual numbers and back numbers can be purchased and a Bulletin catalogue, by series, is available. Orders and enquiries should be sent to:

Sales Department,

Natural History Museum Publications,

British Museum (Natural History),

Cromwell Road,

London SW7 5BD

Telephone: 07 1-938-9386 Telex: 929437 NH PUBS G 07 1-938-8709

World List abbreviation: Bull. Br. Mus. nat. Hist. (Bot.) © British Museum (Natural History), 1990

ISBN 0 565 08026 1 Botany Series

ISSN 0068 - 2292 Vol 20( 1 ), pp 1 - 1 5 1

British Museum (Natural History)

Cromwell Road

London SW7 5BD issued 28 June 1 990

Typeset by Waveney Typesetters, Norwich, Norfolk.

Printed in Great Britain by Henry Ling Ltd., at the Dorset Press, Dorchester, Dorset

Bull. Br. Mus. nat. Hist. (Bot.) 20(1): 1-151

Issued 28 June 1990

Studies in the genus Hypericum L. (Guttiferae) 8. Sections 29. Brathys (part 2) and 30. Trigynobrathys*

17.1111 199

NORMAN K. B. ROBSOJV /

do Department of Botany, British Museum (Natural History), Cromwell Road, London SW7 5BD

PRESENTEC GENERAL

CONTENTS

Synopsis ................... 1

Introduction .................. 2

Separation of sects Brathys and Trigynobrathys ............ 2

Sect. 29. Brathys 3

(a) Characters ................. 4

(b) Distribution and evolution 4

Sect. 30. Trigynobrathys and its subdivision ............ 5

Morphology of sect. 30. Trygynobrathys ............ 6

(a) Leaves 6

(b) Inflorescences 6

Cytology and hybrids in sect. 30. Trigynobrathys ........... 8

Distribution and evolution in sect. 30. Trigynobrathys .......... 8

(a) Principal clades (i) and (ii): relatives of H. rigidum

subsp. sellowianum ................ 8

(b) Principal clade (iii): H. brasiliense, H. campestre, and

relatives .................. 9

(d) Wide disjunctions: discussion 11

Systematic treatment 12

Sect. 29. Brathys (Mut\se\L.f.)Cho\sy 12

Sect. 30. Trigynobrathys (Y. Kimura) N. Robson 47

References 146

Systematic index 148

SYNOPSIS. New collections from Colombia (J. R. I. Wood) and Ecuador (Flora of Ecuador collectors) have necessitated a reorganization of Hypericum sect. 29. Brathys. The mainly woody species, described in Part 7 (Robson, 1987), have increased in number from 61 to 75 by the addition of 11 new species (one as yet unnamed) and the transfer of three species provisionally allocated to sect. Knifa (Robson, 1987). The remainder of this section has been reduced to subsectional status under sect. 30. Trigynobrathys. The treatment of sect. Brathys is completed by accounts of 13 mainly herbaceous species not included in Part 7, two of which are new. A new enumeration of the 88 species of sect. Brathys and a revised key are provided, and the section is divided into four subsections: 1. subsect. Styphelioides N. Robson, subsect. nov., 2. subsect. Phellotes N. Robson, subsect. nov., 3. subsect. Brathys, 4. subsect. Spachium R. Keller. Sect. 30 Trigynobrathys is divided into two subsections: subsect. Connatum (R. Keller) N. Robson, comb. nov. and subsect. Knifa (Adanson) N. Robson, stat.nov., and a systematic account is given of the 52 species. A discussion of the morphology, chromosome numbers, hybrids, distribution, and evolution of each section is included.

The following other new taxa (sp. nov. or subsp. nov.) are described in sect. 29. Brathys: Hypericum phellos subsp. marcescens N. Robson (Colombia: Santander), H. matangense N. Robson (Ecuador: Morona-Santiago), H. callacallanum N. Robson (Peru: Amazonas), H. gladiatum N. Robson (Colombia: Boyaca), H. espinaliiN. Robson (Colombia: Cauca), H. castellanoi N. Robson (Colombia: Boyaca, Venezuela: Merida), H. asplundii N. Robson (Ecuador: Pichincha), H. radicans (Columbia: Boyaca), H. paramitanum N. Robson (Venezuela: Trujillo), H. woodianum N. Robson (Colombia: Cundinamarca), H. harlingii N. Robson (Ecuador: Loja), H. rubritinctum N. Robson (Mexico: Guerrero, Mexico), H. beamanii N. Robson (Guatemala: Solola); in sect. 30. Trigynobrathys: H. cordatum subsp. kleinii N. Robson (Brazil: Parana, Sta Catarina), H. pedersenii N. Robson (Brazil: R. G. do Sul), H. campestre subsp. pauciflorum N. Robson (Brazil: Parana to R. G. do Sul, Argentina: Chaco), H. campestre subsp. tenue N. Robson (Paraguay, Argentina: Misiones), H. salvadorense N. Robson (Brazil: R. G. do Sul), H. silenoides subsp. minus N. Robson (Galapagos Is), H. relictum N. Robson (Colombia: Santander), H. killipii N. Robson (Colombia: Santander, Norte de Santander).

'For part 1 see Robson (1977); for part 2 see Robson (1981); for part 3 see Robson (1985); for part 7 see Robson (1987).

N. K. B. ROBSON

The following changes of rank (stat. nov.) are also made: sect. 29. Brathys - Hyperlcum chamaemyrtus subsp. pseudocaracasanum (Steyerm.) N. Robson (Venezuela: Lara, Trujillo); sect. 30. Trigynobrathys - H. rigidum subsp. meridionale (Lyman B. Smith) N. Robson (Brazil: Sao Paulo to Sta Catarina), H. rigidum subsp. sellowianum (R. Keller) N. Robson (Brazil: Minas Geraes to R. G. do Sul), H. denticulatum subsp. acutifolium (Elliott) N. Robson (south-eastern U.S.A.), H. myrianthum subsp. tamariscinum (Cham. & Schlechtendal) N. Robson (Brazil: R. G. do Sul, Uruguay), H. mutilum subsp. latisepalum (Fern.) N. Robson (south-eastern U.S.A.), H. pumillum subsp. diffusum (Rose) N. Robson (north-central Mexico). One is also a new name (stat. et nom. nov.): H. rigidum subsp. bracteatum N. Robson. One new combination (comb, nov.) is made: H. cordatum (Veil.) N. Robson.

INTRODUCTION

Separation of sects Brathys and Trigynobrathys

One of the main problems in revising the sectional classifica- tion of Hypericum L. has been the recognition of the boundary between Brathys and 'Spachium'. Keller (1893) ranked these taxa as subsections, differentiating them respec- tively as shrubs or subshrubs (subsect. Eubrathys) and herbs (subsect. Spachium). He distinguished two further groups, subsects Connatum (subshrubs with connate leaves - H. connatum Lam.) and Multistamineum (herbs with spike-like inflorescence and numerous stamens - H. angulosum Michaux = H. denticulatum subsp. acutifolium and H. pilosum Michaux = H. setosum), for species that did not fall easily into the major shrub/herb categories. When revising his treatment for the second edition of Die naturlichen Pflanzen- familien (Keller, 1925), he merged these last two subsections with subsect. Spachium and elaborated his differentia:

Subsect. Eubrathys. Subshrubs or small shrubs with often dense appressed foliage; flowers often terminal, usually relatively large, i.e. about 1-5 cm in diameter, rarely less.

Subsect. Spachium. Annual or perennial herbs; stamens 5- 50, rarely more.

Subsect. Eubrathys in Keller's revised treatment still contained two species that are now allocated to other sections (H. nitidum Lam. in sect. 20. Myriandra and H. epigeium R. Keller in sect. 9. Hypericum), as well as several that should have been in his subsect. Spachium. This, in turn, contains some species that would now be allocated to sects 9. Hypericum and 26. Humifusoideum, although most do belong to 'Spachium' .

In the preliminary key to sections in Part 1 (Robson, 1977: 342), I still maintained the 'mainly shrubby/mainly herba- ceous' distinction, supplementing it by characters of inflor- escence, leaf margin, and stem internodes and raising Brathys and Spachium to sectional rank. Further work, however, showed this view to be untenable, because (i) sect. Brathys was found to contain herbaceous species, including the type of sect. Spachium (H. gentianoides (L.) Britton, Sterns & Pogg.) and (ii) the subsidiary characters that were used are not completely diagnostic.

By combining leaf-insertion and leaf-venation characters with type of inflorescence branching, it seemed possible to extract, partly from each section, a group with long narrow leaves and pinnate or single veins. This was described in sect. Knifa (Adans.) N. Robson in Part 7 (Robson, 1987) and the correct name for sect. 'Spachium' excluding H. gentianoides and its relatives was shown to be sect. Trigynobrathys (Y. Kimura) N. Robson.

Scarcely had Part 7 gone to press, however, when a study of new material from Colombia (J. R. I. Wood) and Ecuador (Swedish and Danish collectors from GB and AAU) showed that sect. Knifa in turn was untenable, as it consisted of a shrubby element belonging to sect. Brathys and a herbaceous element belonging to sect. Trigynobrathys. Even then, the dividing line between these two sections was not completely defined and has remained 'fluid' throughout most of the time during which Part 8 has been written. I am now satisfied, however, that all species have been allocated to the appropri- ate section, and the distinctions between the sections have become clear, though still not easy to describe.

Sect. Brathys comprises 88 species of shrubs (rarely small trees), subshrubs, shrublets or wiry annuals (rarely peren- nials) with short stem-internodes, leaves plane to completely revolute or completely involute and often united in pairs but not perfoliate, flowers solitary with inflorescence-branching pseudo-dichotomous or (especially in the less shrubby spe- cies) dichasial/monochasial or mixed, stamen fascicles almost never distinguishable, and styles usually 3 but plesiomorphi- cally 5(4) and sometimes apomorphically (4)5. It has two primary centres: (i) two closely related species in Belize and Cuba and (ii) the Venezuela-Colombia border (see Part 7: 7). From the latter area the mainly shrubby part has radiated eastward to Roraima, westward to Costa Rica, and south- westward along the Andes to Bolivia, with a disjunct species (Hpiriai) in south-east Brazil and Uruguay. The subshrub-to- annual part has its primitive species in northern Colombia and north-west Venezuela and a secondary centre in southern Mexico. Thence it has spread south-eastward to Panama, eastward to the Greater Antilles, north-westward to southern Baja California and the Revilla Gigedo Islands and, with a disjunction, north-eastward to the eastern U.S.A. and south- eastern Canada. The 88 species can now be seen to fall into four distinct groups, which are described as subsections.

Sect. Trigynobrathys comprises 52 species of shrubs, subshrubs, shrublets, perennial and annual herbs with stem internodes short to elongate, leaves plane to recurved or very rarely incurved and free or perfoliate, flowers solitary or more usually in dichasial/monochasial cymes or with bran- ching mixed cymose/pseudo-dichotomous (rarely wholly pseudo-dichotomous), and styles 5-2 with two independent reversals from 3 to 5. The primary area of speciation here has been south-eastern Brazil, with a primary radiation (to the south of the Amazon) northward to eastern Brazil (Bahia), and (with major disjunction) to south-eastern U.S.A., south- ward to Uruguay and northern Argentina, westward to Bolivia, Peru, and northern Chile and thence north along the Andes to southern Colombia and the Galapagos Islands, and west (with disjunction) to New Zealand, Australia, New Caledonia, and New Guinea and (with further disjunction(s)) to scattered localities in south-east Asia. From a secondary centre in north-east Colombia it has spread into eastern Costa

HYPER/CUM

Sect. 29 Brathys

Relationships

wurdackii

castellanoi19 \ 20 asplundii | <huyoides30

gladiatum

sp. 17-16 15

Brathys

60 tetrastichum

sabiniforme martense

espmalii 0 _.

21 radicans

quitense W0odianum .

25 26 29 \ 34

... ,. hartwegu /acicu are

myricarnfplium , nf<; u / /

fcO I / /

lycopodioides /sprucei

co 58 lancifolium

46 bolivaricum valleanum

millefolium

67 45 --57 /

marahuacanum baccharoidesl magdale'hicum x 66 py cnophyllum

A . II 56

4 1 w -parallelum

,« 42 cassiopiforme

37 8tu4er

gleasonii

64 69 struthiolifolium

caracasanum harlingii

pimeleoides

44 62 ,^ selaginella

jaramilloi 68 -r q f A

ruscoides . . rubritinctum78 ^^ lancioides

carinosum 6 J

matangense

arbuscula, , fuertesii 80dich°to

cymobrathys ' ___.77 — — <~~- «

84 Peninsulare diosmoides

ssp. phellos ssp.

marcescens

e aj t w 9 o d i a n u m

86 galinum gentianoides

Fig. 1 Sect. 29 Brathys. Relationships of the 88 species and related sections. Limits of (named) subsections indicated by bold lines. Revised and

expanded from Part 7.

Rica and south along the Andes to northern Argentina and (with disjunction) central Chile. From this secondary centre has arisen a tertiary centre in western Mexico, whence there are trends south-east to Nicaragua and the Greater Antilles, north to the U.S.A. and southern Canada, and east to Cuba. Major disjunctions from the secondary centre have been followed by speciation in Africa and East Asia. Despite these centres of speciation and radiation, only two groups are sufficiently distinct to warrant designation as subsections in this section; and homoplasy (convergence) is so rife as to make a workable key difficult to construct.

As a result of the circumscriptions outlined above, Brathys and Trigynobrathys can be seen to form distinct entities with distinct but multiple overlapping distributional areas; but there are no single characters that can be used to separate them. Even distinctive combinations of characters are dif- ficult to find:

Sect. Brathys: Trees, shrubs, shrublets or wiry herbs. Stem glands sparse or apparently absent; internodes usually shorter than leaves. Leaf pairs free or forming interfoliar ridge or very rarely ± united at base; lamina mostly plane to involute, glands always punctiform. Inflorescence 1 -flowered with branching pseudo-dichotomous or, more rarely, 2-15- flowered with branching dichasial/monochasial or mixed. Styles (5-4)3.

Sect. Trigynobrathys: Shrubs, subshrubs or ± soft herbs.

Stem glands dense; internodes (except basal) usually equal- ling or exceeding leaves. Leaf pairs free or ± united at base, not forming interfoliar ridge; lamina mostly plane to revolute, glands sometimes elongate. Inflorescence 2-°°-flowered with branching dichasial/monochasial to sympodial or mixed or, more rarely, 1-flowered with branching pseudo-dichotomous. Styles 5-3(2).

Sect. 29. Brathys

The revised sect. Brathys of 88 species, not 80 as stated in Part 7, and the revised numbering and suggested relationships are shown in Fig. 1. As before, one species is shown as directly derived from another where its characters are all or nearly all apomorphic in relation to the 'ancestral' one. A comparison with Part 7: 4, fig. 1 (using the revised number- ing) shows that the relationships of 3. H. phellos and Spp. 4- 10 are somewhat simpler than was at first thought; 11. H. simonsii and 12. H. papillosum are basic to the H. humbold- tianum group (Spp. 13-21), not to the H. lycopodioides group (Spp. 23-35), which is related to a new, relict species (22. H. paramitanum) e.g. by the basically pinnate leaf venation. All these form subsect. 2. Phellotes. Subsect. 3. Brathys is as before except for the addition of 39. H. baccharoides and the new 69. H. harlingii; and subsect. 4. Spachium is shown as a

N. K. B. ROBSON

Characters

united-free >' I .' /

partly,'not|whol'ly/ I I ,' I 3 ;4| 5 styles

-*v*"^ -- O / ~.

ipseudopetiolate---^ _ — QQ

leaf [base

i sessile

-86/

Fig. 2 Sect. 29 Brathys. Limits of certain characters within the section and related sections. Note the isolated apomorphic occurrences of 4-5-

styled species. Revised and expanded from Part 7.

sister-group of subsect. Brathys alone rather than of subsects Phellotes + Brathys.

(a) Characters

The revised character diagram (Fig. 2) shows the limits of the same characters as were depicted in Part 7: 6, fig. 2. Apart from the extensions to subsect. Spachium, these diagrams are basically similar. No more trend reversals (e.g. in Spp. 54-55, 67, and 74) have been revealed, except possibly the partly deciduous leaves in Spp. 4 and 5. Indeed, what appeared to be a reversal to wholly persistent leaves in Spp. 9-10 is now seen not to be one; instead the partly deciduous leaves of Sp. 8 are shown to be a specialization. One additional character in the diagram (leaves free v. united) nicely delimits subsect. Brathys from subsects Styphelioides and Spachium, but sub- sect. Phellotes contains species with both conditions. Finally, the pseudopetiolate leaf base in subsect. Spachium (Spp. 75- 76) is seen to have arisen independently from that in subsect. Brathys.

(b) Distribution and evolution (Fig. 3)

The six morphological group-headings (a-f) under which the distribution and evolution of sect. Brathys were considered in

Part 7: 7-12 have been reduced to three subsections, and a fourth (subsect. Spachium) has been added. Subsect. 1. Styphelioides is confined to Belize and Cuba. Subsect. 2. Phellotes is based (morphologically and geographically) in the Colombia/Venezuela border area and extends eastward to Guyana (Roraima), northward to Costa Rica and adjacent Panama, and south-westward along the Andes to northern Peru and, with disjunction, south-eastern Brazil and Uruguay. Subsect. 3. Brathys has the same centre as subsect. 2 and a similar distribution; it occurs also in Hispaniola and extends along the Andes as far south as Bolivia, but it does not occur in south Brazil. It is present as well in Roraima (but only the Venezuelan part) and also in southern Venezuela (Cerro de Marahuaca).

Subsect. Spachium, too, radiates from north-eastern Col- ombia (75. H. cymobrathys), but mainly northward. Apart from 76. H. chamaemyrtus , with a Colombian and a Vene- zuelan subspecies, all the other species are natives of Central and North America and the West Indies. Disjunctions between northern South America (76. H. chamaemyrtus) on the one hand, and southern Mexico (77. H. arbuscula) and Honduras (primitive form of 82. H. gnidioides) on the other, suggest a northward migration through the early, Greater Antilles land-link between the two Americas (Part 7: 9), with subsequent migration and differentiation from the Belize/

HYPER/CUM

Sect. 29 Brathys

C?£W) ?21CO(nc) E(s)-PE(n)

•V. 27

^ C0(c)

25 26

24 _J c°(nc)

CO (c)

Distribution

CO(nc)-V(w)

C0(he)

E(nh CR'-P'A cocsw}-

E(n>PE(n)43

PE(n)42

"w 53 58co(r'<0-v(w) c o(wc) N / c ° (ny 5 7 b v W-c °(n e)

67HA

CO(c)-VQw) C0(n)-V(nw) E (s)

C0(ne)-V(w)

CO(nne)\Co(-ne)V(.v

4 +- 3c— 3b=3a

CR-PA(w) CO(ne)

37 40

co PE(n)x

(ne - nc)

CO(nc) -V(w)

PE(s) BO

DO V/ n 7 n

CO(ne-nw)XD /U 7 1 E^

+CR , 64 69 I >/

! »^ PE(n-s)

CO(ne-nc) 44 — 62+CR C0(nc-c)

C0(ne-nc) "68— 7 o _ -7 A

COCc)-E(c) 'O ^

'.O E(n)-V(w)

', HAjPRjcU

76a

^30| C0(nc-c)\

USA(e),CAN(se)

Fig. 3 Sect. 29 Brathys. Distribution of the 88 species showing major and minor disjunctions (= across sea, across land (major). across

land (minor)). Lower-case letters indicate compass points. Revised and expanded from Part 7. For geographical abbreviations see p. 9.

Guatemala region. The more primitive species, with longer styles (77. H. arbuscula), contains an indivisible morphocline from southern Mexico (Chiapas) to Guatemala, whence there is a disjunction in this clade to Hispaniola (79. H. fuertesii, 80. H. dichotomum, and 81. H. diosmoides). These species form an increasingly herbaceous and procumbent series, and the most specialized one, H. diosmoides, has extended its area into Puerto Rica and Cuba. Whether the disjunction between Guatemala and Hispaniola is a result of overland or long- distance migration is unclear. The latter seems more likely in view of the relatively specialized form of these species. Northward, //. arbuscula or its predecessor has given rise to 78. H. rubritinctum, which occurs in central Mexico (Guer- rero, Mexico).

The initially more southern branch of the Central Amer- ican radiation starts with 82. H. gnidioides, which is very variable in Honduras and, to a lesser extent, Nicaragua, and shows a south-eastward morphocline to western Panama. Here there occurs an extremely reduced prostrate form ('H. woodsonii'), which is linked to the more typical one by a continuous series of intermediates. North-west from Hon- duras the nearest relatives of H. gnidioides are (a) in west- central Mexico and the Revilla Gigedo Islands (83. H. eastwoodianum) with a derivative species in the extreme south of the Baja California peninsula (84. H. peninsulare)

and (b) in east-central Mexico (86. H. galinum) with a sister- species in the mountains of Guatemala (85. H. beamanii). Finally, 87. H. drummondii , from south-eastern U.S.A., is morphologically very close to the least shrubby (annual?) Honduran form of 82. H. gnidioides; and 88. H. gentianoides (eastern U.S.A. and extreme southern Canada) is more reduced than H. drummondii. There is no evidence, in the form of relict species (etc.), that H. drummondii has reached the southern United States overland. Indeed, its primitive form is so close morphologically to H. gnidioides that such a long migration without greater differentiation seems unlikely. I am therefore inclined to attribute the disjunction to ancient long-distance migration, possibly the result of carriage of propagules by extra-strong winds or birds.

Sect. 30. Trigynobrathys and its subdivision

As here delimited, sect. Trigynobrathys comprises 52 species that vary widely in habit from shrubs 2 m tall to prostrate annual herbs (Fig. 4). It can be divided into two subdistinct groups:

1) Shrubs, subshrubs or woody annuals; styles 5(4-3); capsule usually broadly ovoid to globose; mainly south of the Amazon, also Chile, the Andes from Peru to Colombia,

N. K. B. ROBSON

Sect. 30 Trigynobrathys

Relationships

22b

ssp. tamariscinum

salvadorense

17 polyanthemum

anceps 24

25 b ssp. minus

26 caespitosum

35 aphyllum

25a

ssp. silenoides

22a

ssp myrianthum

gramineum

16, 14

31 brevistylum

lorentzianum 21

brasiliense 24?

30,

thesiifolium

ssp. campestre 18b

paucif lorurr

18a ssp.pauciflorum

ssp.acutifolium 24 10b

\ •*

\ gymnanthum38 re

denticulatum

cumulicola

10a

ssp. denticulatum

24,48

japonicum re

ssp.diff usum

43a pumillum

43b ssp.

pumillum

ssp. bracteatum 1 d

1b SSP meridionale

ssp. sellowianum

anagalloides re 49

globulif erum

microlicioides Brathysk 29

ssp. rigidum

scioanum re Knifa

legrandii 9

ssp.kleinii

cavernicola

connatum

caprifoliatum

Campylosporus I * "*

arenarioides 40

37 /

canadense ssp./

16 latisepalum

39b

3ga mutilum

ssp.mutilum re ^^

39c

ssp. boreale ref is)

pleiostylum

45 paucifolium

moranense

oligandrum

Fig. 4 Sect. 30 Trigynobrathys. Relationships of the 52 species and related sections, showing known diploid chromosome numbers. Limits of

(named) subsections indicated by bold line. Subspecies linked by paired lines.

(a) Leaves

Galapagos Is., eastern N. America, and Australasia north to Taiwan, Vietnam, and Bhutan ('Trigynobrathys proper'). Spp. 1-27.

2) Subshrubs, shrublets or herbaceous annuals; styles (2) 3(4- 8); capsule usually narrowly ovoid to ellipsoid or cylindric; mainly north of the Amazon, Andes from Venezuela south to Argentina, Central and North America and Greater Antilles, also Africa, Madagascar, and east Asia, south Australia, and New Zealand ('Knifa'). Spp. 28-52.

The presence of 25. H. silenoides, 26. H. caespitosum, and 27. H. gramineum in group 1) and of various other exceptions to the 'broad 5-styled ovary v. narrow 3-styled ovary' division makes it difficult to define Knifa as a subsection. An undivided section of 52 species, however, is rather unwieldy and so an attempt has been made to do so as the subsections in practice are usually easily distinguishable (see pp. 51 and 95).

Morphology of sect. 30. Trigynobrathys (Fig. 5)

The species in this section vary from tall shrubs (up to 2 m) to erect or prostrate herbaceous annuals. They all lack dark glands completely and have no resin glands on the lower surface of the leaf; and there are no fringing glands on leaves, sepals or petals. All species have marginal punctiform leaf glands (often very small). The laminar glands are usually also wholly punctiform, but very rarely lines of punctate glands may coalesce to produce streaks (30. H. thesiifolium in part).

As in sect. 29. Brathys, the leaves in part of the species with the most primitive characters (1. H. rigidum) have parallel and open venation; but even within this species, forms with partially pinnate venation occur. Pinnation occurs in two ways:

i) by union of the laterals with the midrib, the outer ones nearest the base, with the venation remaining apparently open (e.g. in 6. H. cordatum, Fig. 13) or becoming closed (e.g. in 4. H. caprifoliatum).

ii) by development of secondary laterals along the midrib as the space between it and the primary laterals increases (e.g. in the most primitive form of 16. H. brasiliense). Method i) is often combined with method ii), as in 18. H. campestre, Fig. 15 A-C. Other primary laterals may produce secondaries on the marginal side, and the system may become close (e.g. in 25. H. silenoides, Fig. 17 A-B).

In the more shrubby species with free leaf-bases, the leaves are sooner or later deciduous (e.g. in 14. H. denudatum); but where the leaves are herbaceous or the bases united, they persist. In no species are they partially deciduous, as they are in many species of sect. Brathys. Again, in contrast to much of the sect. Brathys, the lamina is usually plane to revolute, rarely involute or conduplicate (e.g. in 9. H. legrandii or 13. H. cumulicola).

(b) Inflorescences

Unlike the species of sect. Brathys, where the 1 -flowered

HYPERICUM

Sect. 30 Trigynobrathys

Characters

39c

perennial or """•""•• 5K3(D shrub\subshrub \ shrublet .--'"'annual

Fig. 5 Sect. 30 Trigynobrathys. Limits of certain characters within the section and related sections. Note the isolated apomorphic occurrences of

4-5-styled species.

inflorescence with pseudo-dichotomous branching is com- moner than the dichasium/monochasium, those of sect. Trigynobrathys more frequently have dichasial/monochasial branching, which can become elaborate (e.g. to the sixth grade in 22. H. myrianthum, Fig. 16 B-C). Pseudo- dichotomous branching, however, is far from absent (e.g. in 50. H. humbertii, Fig. 22 A), and mixed inflorescences with both types of elaboration are of frequent occurrence in subsect. Knifa. In some advanced species of this subsection, only one branch of the pseudo-dichotomy normally develops, resulting in a sympodium (e.g. in 42. H. japonicum variant 5. 'humifusum').

Lateral inflorescences are usually present as well as the terminal one, but they sometimes are slow to develop (e.g. in 27. H. gramineum, Fig. 17 E).

Variants of 1. H. rigidum, the species most similar to those of the African sect. 1. Campy losporus, already display most combinations of these types of inflorescence elaboration: 1- flowered with branching wholly lateral (Fig. 12 C), 1-3- flowered with branching wholly lateral (Fig. 12 B), 1-7- flowered with branching at first pseudo-dichotomous then lateral (Fig. 12 A).

As is usual in sect. Brathys, the perianth in sect. Trigyno- brathys is pentamerous with the sepals appressed to the

petals. These are usually longer than the sepals and are always spreading ('flowers stellate'). The sepals vary with the relative width of the base from 1-9-nerved, the nerves often becoming (more) prominent in fruit. There are trends in the petals from presence to absence of an apiculus and pellucid glands, but these are of no more than specific taxonomic value. As in sect. Brathys, the petals and stamens are persistent, the petals rolling downward after anthesis and sometimes enclosing the developed capsule.

The number and size of the stamens are correlated in general with flower size as a whole. In the larger flowers the five fascicles are sometimes distinct at the base (e.g. in 16. //. brasiliense; c.f. Reichardt, 1878: fig. 34), but they usually form a narrow complete whorl. In the smallest flowers (e.g. in 37. H. canadense or 51. H. scioanum) the reduced number of stamens separates again into groups, normally three but sometimes five or an irregular grouping that does not correspond to that of the original fascicles.

The styles vary from 5 to 3 or rarely 2 (e.g. in forms of 42. H. japonicum). The intermediate number 4, while never uniquely characteristic of a species, is possibly commoner in sect. Trigynobrathys subsect. Connatum than in most other sections of Hypericum. li recurs in several species, nearly all in subsect. Knifa, which are normally 3-styled. As was stated above, however, there are two species (41. H. pleiostylum, 52. H. oligandrum) in which this trend is further reversed, the number of styles increasing in H. oligandrum to 4-5 and in H.

N. K. B. ROBSON

pleiostylum up to 8. The form and behaviour of the styles do not appear to have taxonomic value, although there is some variation: most curve outward, but some later curve inward again. The form of the stigmas, however, can sometimes be utilized; it varies from not or scarcely enlarged (rare) to very broadly capitate or even peltate, and the style immediately below is often gradually enlarged. This gradual enlargement may occur below a narrow, even rounded, apex (stigma clavate), a state characteristic of the H. lalandii group (Spp. 48-52).

Cytology and hybrids in sect. 30. Trigynobrathys

The most primitive chromosome number in the genus, n = 12, occurs in subsect. Connatum, in 10. H. denticulatum , 11. H. harperi, and possibly 16. H. brasiliense, and the only polyploid number so far recorded (n = 24) is found in the H. denticulatum group where H. denticulatum subsp. denticula- tum is sometimes diploid, sometimes (auto-?) tetraploid. The two more advanced members of this group, however, have n = 6. This number cannot (as Webb (1980) averred) here represent the basic number of the whole genus, thus making n = 12 tetraploid. The species with this number, 12. H. setosum and 13. H. cumulicola, are clearly derived from 10. H. denti- culatum (n = 12, 24) and, for whatever reason, must be regarded as haploids unless (which seems more unlikely) the whole complement is still present with the chromosomes united in pairs.

The numbers n — 11-9 have not yet been recorded in sect. Trigynobrathys except for n = 9 in 39c. H. mutilum subsp. boreale (Kapoor, 1972), where it represents a secondary increase in a taxon that normally has n = 8. Otherwise all chromosome counts of species in subsect. Knifa and else- where in subsect. Connata have yielded n = 8 or 2n = 16, the sole exception being a New Guinea plant of H. gramineum (n = 7) (Borgmann, 1964, as Hypericum sp.). Whether this gap in numbers is real or, more likely, merely the result of insufficient work, remains to be seen.

Hybrids have been found only in the H. canadense group (Spp. 36-42), all the North American (i.e. excluding Mex- ican) species of which are apparently interfertile. Webb (1980) reported that hybrids and putative hybrids in this group showed a lower chromosome stainability than did species. From field observations and herbarium studies, he deduced that the name H. dissimulatum Bickn. has been applied to recurrent hybrids between 37. H. canadense and two subspecies of 39. H. mutilum (subsp. mutilum and subsp. boreale} .

The only other suspected hybrid that has been reported is from Bhutan, where the distributions of 27. H. gramineum and 42. H. japonicum variant 4c ('calyculatum') overlap (Robson, 1972: 267).

Distribution and evolution in sect. 30. Trigynobrathys (Fig. 6)

Whereas sect. Brathys is centred north of the Amazon, in the north-eastern Andes and Greater Antilles, sect. Trigynob- rathys has its centre in southern Brazil. The evolutionary divergence pattern of Hypericum as a whole suggests that the split between them occurred early, probably in Africa before the progenitors of either section reached South America (Robson, 1987: 7). The transference of the shrubby species

from (subsect.) Knifa to sect. Brathys makes it seem even less likely that there was a later sectional divergence across the Amazon basin. Indeed, the disjunction of subsect. Knifa from subsect. Connatum, which must of course have been later, involves that very region.

(a) Principal clades (i) and (ii): relatives o/H. rigidum subsp. sellowianum

The most primitive species in sect. Trigynobrathys (i.e. the one that is morphologically nearest to the east African H. revolutum subsp. keniense} is 1. H. rigidum, of which the most primitive subspecies (subsp. a. rigidum) is confined to Parana in southern Brazil and has a closely related derivative species, 2. H. microlicioides, in the same province and also in Santa Catarina. All the other parts of the section are related to other subspecies of H. rigidum. Subsp. b. meridionale, with a somewhat wider distribution (Sao Paulo to Santa Catarina), has apparently given rise to two further subspecies: c. subsp. sellowianum, which has the widest distribution (southern Minas Geraes to Rio Grande do Sul), and d. subsp. bracteatum, which has a relict distribution in Parana.

Of the two major clades related to subsp. sellowianum, one, clade (i), is confined to south-eastern Brazil and adjacent regions of Paraguay, Argentina, and Uruguay. In it the tendency to shorten and broaden the leaves in that subspecies is at first elaborated, so that the leaves are very broad in 3. H. teretiusculum and the pairs connate in 4. H. caprifoliatum and 5. H. connatum. In the other derivatives of H. teretiusculum they become smaller and narrower, and this trend finishes in 9. H. legrandii (Uruguay), where they are linear-incurved. In the related 8. H. ternum- they are sometimes in whorls of three.

The other major clade related to subsp. sellowianum (ii) begins morphologically not far from H. teretiusculum, but geographically its members are very distant - in south-eastern U.S.A. The affinities of the H. denticulatum (Spp. 10-13) are clearly with the H. teretiusculum group (e.g. they have similar leaves and inflorescence-branching), but both groups seem to be related independently to H. rigidum subsp. sellowianum. Although the American species are woody herbs rather than subshrubs, the morphological differences are not great, which suggests that their ancestor reached the U.S.A. by long- distance dispersal (not overland), from south of Amazonia. H. rigidum and H. teretiusculum are plants of moist habitats, and it is not impossible to imagine that seeds of these or of a related species were transported externally (?) by birds on a northward migration. The suggestion of a long-distance origin of the H. denticulatum group invites a comparison with relatively few examples - 'a remarkable group of 6 species disjunct between the eastern United States and temperate South America' (Raven, 1963). These include Hypericum gentianoides (sect. Brathys}, which must have travelled in the opposite direction more recently, being one of the most highly evolved species in that section and not morphologically distinct in South America (c.f. Smith, 1962; Thorne, 1972). Smith (1962) regards Hypericum as one 'of the best examples ... of a wide-ranging Andean element in southern Brazil'; but this scenario applies only to the H. carinatumlH. silenoides and (possibly) H. relictum groups (Spp. 23-26, 28- 29), the relationships of the other southern Brazilian species being more varied.

Examples of birds that regularly migrate from southern Brazil to the south-eastern United States and back have not

HYPER1CUM

Sect. 30 Trigynobrathys

Distribution

BR(s)20

BR(s)17 UR

BR(s)19 UR

26 CH(c)

27 NZ.AUS.NC, HW, ASIA(e-se)

USA(se).

BR(s),UR A R (h e - e , n w)

PY.AR(ne) 18c

7 ' UR

BR(se-s)16 PY,BO,AR(ne,nw;

CO(sw)-BO / AR(nw)

30/ CO,V,PA,CR

ME(se)-HO 34

ME(s)-NI(n)

BE 35

USA(rO,CAN(s) [GE,F, J J 36

USA(e),CAN(e)

[IR, NE.F] NW 39b

^ USA(se) USA(se)38 393

[GTPOjj^ USA(e),CAN(ef [AZ,EUR,SU(sw), NZ.HW.ME.HO, D,CO,E,PE,PY, BR(se)V] USA(ne)

41 BR (se)

BR(s),UR PY,AR(n)

CAN (e)

/ ME(nw) / USA(w) / CAN(sw)

AFR(nw,ne-se,3w) AFR(e) MA

APR (ne-e)

44 ME(c-ne; USA (s)

AFR(sw)

Fig. 6 Sect. 30 Trigynobrathys. Distribution of the 52 species showing major and minor disjunctions (= across sea, — across land). Paired relationship lines indicate subspecies. Lower-case letters indicate compass points. Introductions (ancient or recent) within square brackets.

Geographical abbreviations used in Figs 3 and 6

APR

ASIA

AUS

CAN

Africa

Argentina

Asia

Australia

Azores

Belize

Bolivia

Brazil

Canada

Chile

CO Colombia

CR Costa Rica

CU Cuba

D Dominican Republic

E Ecuador

EUR Europe

F France

GA Galapagos Islands

GE Germany

GT Guatemala

GY Guyana

HA Haiti

HO Honduras

HW Hawaii

IR Ireland

J Japan

MA Madagascar

ME Mexico

NC New Caledonia

NE Netherlands

NI Nicaragua

NZ New Zealand

PA Panama

PE Peru

PR Porto Rico

PY Paraguay

SU Soviet Union

UR Uruguay

USA United States

V Venezuela

yet been found;2 but I have little doubt that such migrations occur. At any rate, the bird(s) that hypothetically transported H. gentianoides in relatively recent times would be unlikely to belong to the same species as the ones that were responsible for the ancient conveyance of the H. denticulatum ancestor. Raven (1963) states that all the temperate N. America/S. America disjuncts date from the late Mesozoic or later, and no more accurate dating seems to be possible yet.

(b) Principal clade (Hi): H. brasiliense, H. campestre, and relatives

The other two principal clades in sect. Trigynobrathys both

2But see Cruden (1966) for a detailed discussion of bird migrants as vectors between California and Chile.

derive from Id. H. rigidum subsp. bracteatum, which is represented as yet only by the type specimen, from southern Brazil (Sao Paulo). Despite its apparent rarity, this sub- species forms a morphological link between the basic species, H. rigidum, and the two remaining principle clades. Of these, (iii) is distributed largely to the south of Amazonia and across the southern Pacific Ocean, whereas clade (iv) (subsect. Knifa) is centred to the north of Amazonia and across the northern Pacific Ocean and the Atlantic Ocean (northern and southern).

Clade (iii) divides into the H. brasiliense group (clade (v), Spp. 14-17) and the H. campestre group (clade (vi), Spp. 18- 27). Clade (v) is confined to south Brazil, Uruguay, and adjacent Argentina, except for 16. H. brasiliense itself, which extends from Parana northward along the eastern hills, Serra

N. K. B. ROBSON

da Mantiqueira and S. do Espinhac,o, to Bahia and westward to Paraguay, southern Bolivia, and northern Argentina and possibly into Peru (see p. 87). Clade (vi), based on the variable 18. H. campestre, is also largely restricted to southern Brazil, Uruguay, and adjacent Argentina, but 18c. H. campestre subsp. tenue occurs in Paraguay and north- eastern Argentina. 23. H. carinatum is more widespread, extending south-westward (in Argentina) to Buenos Aires and Cordoba and westward to Paraguay. In Paraguay it is represented by two forms, one tall and more shrubby ('//. altissimutri , H. megapotamicum') and the other short and more herbaceous, even possibly annual ('//. paraguense'). There are intermediate states elsewhere in the range of the species that preclude taxonomic recognition of these forms, but '//. paraguense' links typical H. carinatum with the Andean and trans-Pacific herbaceous species.

The most primitive form of 25. H. silenoides is in northern coastal Chile ('//. paposum'}, whence there is a northward morphocline along the coastal foothills to Lima. There is a gap in distribution, though scarcely in variation, between there and the Ecuador/Peru border, where a secondary, higher-altitude, cline originates. This extends northward to Colombia (Antioquia) and southward to Bolivia and north- western Argentina. 26. H. caespitosum, from central Chile, is most similar to to some central Peruvian (Ayacucho) forms of this cline, whereas the endemic Galapagos Is. 25b. H. silenoides subsp. minus is morphologically closest to plants of the type subspecies from southern Ecuador. Like (all?) the other Galapagos biota, therefore, it seems to have reached the islands across ocean, probably in this case by means of wind-borne or bird-borne propagules.

The north Chilean form of H. silenoides ('//. paposum') is the most similar part of this species to 27. H. gramineum, which is separated from it by the width of the Pacific Ocean (Robson, 1972: fig. 3 modified). The primitive form of H. gramineum, in turn, occurs in the south of its area (New Zealand, Tasmania, South Australia), whence there are two reduction trends: (a) within Australia northward and west- ward, to small erect annuals and (b) from New Zealand through New Caledonia to New Guinea, to elongate, narrow-