of the angiosperms A revised system of classification R. M. T. DAHLGREN

Transcription

1 BotanicnlJournal offhr Linnenn Sonet?, 80: With 1 figures Fcbi-u;ii.\ A revised system of classification of the angiosperms R. M. T. DAHLGREN Botanical Museum ofthe C niversity of Copenhagen, Gothersgade 130, DK 1123 Copenhagen, Denmark A [cr,b/p(i f.r publication Orcernber / 979 A new \vstetii of classification of the angiosperms is presented. Like that of Dahlgren it is \Iio\vti n; a tw(i-diinensional diagram or frame-work, but differs in numerous major and minor lenturcs. In the tiew classification the angiosperms represent a class divisible into two subclasses cor- I-csponditig to the di- and monocotyledons. The dicotyledons are divided into 24 ruperorders, viz. tlir Magnoliiflorae, Nynphaeiflorae, Ranunculiflorae, Caryophylliflorae, Polygoniflorae, Mnlvilloi-ae, Violiflorae, Theiflorae, Prirnuliflorae, Rosiflorae, Podostemoniflorae, FabiHorae, Proteillorae, Myrtiflorae, Rutiflorae, Santaliflorae, Balanophoriflorae, Araliiflorae, Asteriflorar, Solmifioi-ae, Corniflorae, LoasiHorae, Gentianiflorae and Lamiiflorae, while the monocotyledons r~t-c divided into reven superorders, the Alismatiflorae, Ariflorae, Triuridiflorae. Liliiflorae, Zirigil)c~i-illorae, C:oniineliniflorae and Areciflorae. The higher categories are not defined here, but \oinc general teatures in the classification are mentioned, and the distribution of some attributes is \Iiown in the diagram. KEY WORDS :- angiosperm classification - Liliidae - Magnoliidae - Magnoliopsida. CONTENTS 1111 t-otitrctioii Y1 The classification, Ihc family kvcl Sonic fcnturc-\ ofthe clasihcation The diagratn. possibilities and limitations Distribution ofsoine attributes in thediagram S\-niprtalous flowers Liiitegrnic ovulr\ Presence of parietal cells Endoapn-in forination I 12 Soine c-hemical coiripounds Tlre moii~~cot~lrdotis Pi-ospectives, I-elation to other current classihcations Rclet-enrcs I22 INTRODUCTION A provisional system of classification of the angiosperms was presented in 1975 (Dahlgren, 1975a). A diagram showing the angiosperm orders as figures ( bubbles ) was given in which the relative numbers of species involved were /80/ /$02.00/ The Ltnnean Socier\ of London I

2 92 R. M. T. DAHLGREN indicated by the size of the figures. Their relative positions approximated to the mutual similarity of attributes reflecting phylogenetical affinity. This diagram has been used in a number of publications cited below to illustrate the distribution ofvarious attributes : apocarpy, syncarpy and monocarpellate condition (Dahlgren, 197 7a), choripetaly, sympetaly (Dahlgren, 1977a, b), multistaminate gynoecia with different developmental sequence (Dahlgren, 1977a1, different types of sieve element plastids (Behnke& Dahlgren, 19761, monoaperturate pollen grains (Dahlgren, 1977a, b), sclereid idioblasts (Rao& Das, 1979), types of microsporogenesis (Dahlgren, 1977a), bi- and unitegmicovules (Dahlgren, , tenui-, pseudocrassi- and crassinucellate ovules (Dahlgren, 1975b), types of endosperm formation (Dahlgren, ) bi-and trinucleate pollen grains (Dahlgren, 1975b) iridoid compounds (Jensen, Nielsen & Dahlgren, 19751, ellagic acid and ellagitannins (Dahlgren, 1977a, b) benzylisoquinoline alkaloids (Dahlgren, 1977a, b) tropane alkaloids (Romeike, 1978 pyrrolizidine alkaloids (Culvenor, 1978) polyacetylenes (Dahlgren, 1977a, b) various classes offlavonoid compounds (Gornall, Bohm & Dahlgren, 1979). Other attributes, the distributions of which have been plotted in this diagram, have not been published, such as the seed coat characters presented by Corner (1976). More than a hundred different character states have been plotted in a diagram showing the monocotyledons and the most closely allied dicotyledons (Dahlgren & Clifford, in press), some material of which will be shown in this article. These results have, as expected, shown that my provisional classification and diagram of 1975 could be improved in various respects. Numerous studies on limited groups or on various attributes have contributed to their taxonomic position, such as for the Fouquieriaceae (Dahlgren, Jensen & Nielsen, 1976) and the Retziaceae (Dahlgren, Goldblatt, Nielsen & Rourke, 1979). In the construction of the classification given here, an attempt has been made to take into consideration as much information as possible. The new classification is still provisional, but it is appropriate to present it here in its most updated form. The purpose that can be achieved in the long run should be a system of classification as natural as possible in the sense that it reflects, in the position of each family, phylogenetic relationships. If this is possible, there is a greater chance for predicting the occurrence of an attribute of interest (whether it be the possession of a chemical compound or an embryological feature) or for predicting that a particular, perhaps concealed, attribute is more likely to be found in certain families or orders than in others. Thus a useful system of classification would have a predictive value. This must not be misinterpreted as meaning that all groups with a singular particular property, for example the synthesis of a group of chemical

3 ANGIOSPERM CLASSIFICATION 9 3 compounds, are necessarily related and thus should be in the same part of the classificatory table or diagram, because the same attributes may have evolved independently on separate occasions. Closely allied groups of plants may be very variable with respect to a great number of characters. However, there is little doubt that closely related plants-when a sufficient number of their characters has been examined-will normally prove to agree in more of these characters than will distantly related plants. In order that a classification can be useful for the sake of prediction, it must also be complemented with diagrams showing the approximate distributions of a considerable number of attributes. THE CLASSIFICATION In this classification the angiosperms comprise a class, Magnoliopsida, equivalent to the main groups of the gymnosperms (Pinopsida, Ginkgoopsida, Cycadopsida, Gnetopsida). Their division into two main groups, the dicotyledons and monocotyledons, has often been considered as somewhat arbitrary and would probably not stand a test according to cladistic methods (see Bremer & Wanntorp, 19771, but the monocotyledons stand out as a possibly monophyletic group by virtue of their single cotyledon and characteristic triangular protein bodies in the sieve tube plastids (Behnke, 1969), in addition to a number of less constant differences from the dicotyledons. The dicotyledons and monocotyledons have here been given the rank of subclasses comprising the Magnoliideae and Liliidae respectively, these entities having thus a circumscription different from, and much broader than, the classifications of Cronquist (1968) and Takhtajan (1969). The dicotyledons and monocotyledons are further divided into superorders. In accordance with Thorne (1968, 1976) these bear the termination -JEorae which takes precedence over the -anae used by Cronquist (1968) Takhtajan (1969) and Dahlgren (1975a). A disadvantage with this suffix is that it cannot be used for superorders of lower plants. This classification has been worked out gradually in discussions with colleagues in Europe and North America over a period of years. Rather than aiming to be conservative it is a radical approach, introducing changes where reasonable evidence is available. The farnib level With regard to the Liliiflorae a splitter s approach has been chosen at the family level, largely in accordance with the work of Huber ( 1969). This may seem infatory for a traditionalist used to the Liliaceae sensu lato, and an intermediate level for the family rank in this group may be feasible or even preferable (p. 106). What is important in this context are the basic concepts of interrelationships between the families in the orders presented here, uiz. the Liliales, Asparagales, Dioscoreales, etc. Thus Alliaceae and Hyacinthaceae representing the subfamilies Allioideae and Scilloideae respectively of the traditional Liliaceae are very far from Liliaceae sensu stricto (subfam. Liliodeae), and much closer to Amaryllidaceae, whereas Colchicaceae (Liliaceae subfam. Colchicoideae and relatives) and Alstroemeriaceae are indeed close to Liliaceae sensu stricto (subfam. Lilioideae).

4 94 R. M. T. DAHLGREN The background for these considerations is explained by Huber (1969) and also by the author (Dahlgren et al., 1976) and will be further elucidated in a coming textbook on monocotyledons (Dahlgren & Clifford, in press); the basis is also partly expressed in my previous classification (Dahlgren, 1975a). A similar approach is preferred here for the derivatives of the traditional Saxitragaceae sensu lato. This was very heterogenous and consisted of distantly related elements, which mainly fall into two orders, Saxifragales and Cornales, though some incidently fall outside these orders (e.g., Parnassiaceae, Ribesiaceae). Only a thorough study of the small groups recognized as separate families here and placed for the most part in Cornales, can correctly show the affinities and most appropriate status and circumscription of each. CLASS MAGNOLIOPSIDA (=ANGIOSPERMAE) SUBCLASS MAGNOLIIDAE (=DICOTYLEDONEAE) MAGNOLIIFLO RAE ANNONALES : Annonaceae, Myristicaceae, Eupomaticeae, Canellaceae. ARISTOLOCHIALES: Aristolochiaceae. RAFFLESIALES : RafHesiaceae (including Cytinaceae and Mitrastemonaceae), H ydnoraceae. MAGNOLIALES: Winteraceae, Degeneriaceae, Himantandraceae, Magnoliaceae, Lactoridaceae, Chloranthaceae, I LLICIALES : I lliciaceae, Schisandraceae. LAURALES: Amborellaceae, Austrobaileyaceae, Trimeniaceae, Monimiaceae (including Siparunaceae and Atherospermataceae), Gomortegaceae, Calycanthaceae (including Idiospermaceae), Lauraceae, Hernandiaceae (including Gyrocarpaceae). NELUMBONALES : Nelumbonaceae. NYMPHAEIFLORAE PIPERALES : Saururaceae, Piperaceae (including Peperomiaceae). NYMPHAEALES: Cabombaceae, Ceratophyllaceae, Nymphaeaceae (including Barclayaceae). RANUNCULIFLORAE RANUNCULALES: Lardizabalaceae, Sargentodoxaceae, Menispermaceae, Kingdoniaceae, Circaeasteraceae, Ranunculaceae (including Hydrastidaceae), Berberidaceae (including Glaucidiaceae, Leonticaceae and Podophyllaceae), Nandinaceae. PAPAVERALES : Papaveraceae, Fumariaceae (including Hypecoaceae). CARYOPHYLLIFLO RAE CARYOPHYLLALES: Phytolaccaceae (including Achatocarpaceae, Agdestidaceae and Limeum), Basellaceae, Portulacaceae, Stegnospermataceae, Nyctaginaceae, Aizoaceae (including Mesembryanthemaceae and Tetragoniaceae), Didiereaceae, Cactaceae, Hectorellaceae, Halophytaceae, Chenopodiaceae (including Dysphaniaceae), Amaranthaceae, Molluginaceae, Caryophyllaceae (including Illecebraceae). POLYGONIFLORAE POLYGO NALES : Polygonaceae.

5 ANGIOSPERM CLASSIFICATION 9' MALVIFLORAE (= Dilleniiflorae) PAEONIALES: Paeoniaceae. DILLENIALES : Dilleniaceae. MALVALES: Sterculiaceae, Elaeocarpaceae, Plagiopteraceae, Bixaceae, Cochlospermaceae, Cistaceae, Sphaerosepalaceae, Sarcolaenaceae, Huaceae, Tiliaceae, Dipterocarpaceae, Malvaceae, Bombacaceae. URTICALES: Ulmaceae, Moraceae, Cecropiaceae, Barbeyaceae, Cannabinaceae, Urticaceae. RHAMNALES: Rhamnaceae. ELAEACNALES : Elaeagnaceae. PLUMBAGINALES (position uncertain) : Limoniaceae, Plumbaginaceae. THYM ELAEALES: Thymelaeaceae. EUPHORBIALES: Euphorbiaceae (including Picrodendraceae, Hymenocardiaceae and Uapacaceae), Simmondsiaceae, Pandaceae, Aextoxicaceae (position uncertain), Dichapetalaceae, Didymelaceae (position uncertain; alternativelv in Buxales). VIOLIFLORAE VIOLALES: Flacourtiaceae (including Lacistemaceae), Passifloraceae, Dipentodontaceae, Peridiscaceae (position uncertain), Scyphostegiaceae, Violaceae, Turneraceae, Malesherbiaceae, Achariaceae, Datiscaceae, Begoniaceae, Cucurbitaceae, Caricaceae. SALICALES: Salicaceae. TAMARI CALES : Tamaricaceae, Frankeniaceae. CAPPARALES: Capparaceae (including Cleomaceae, Pentadiplandraceae and Koeberliniaceae), Brassicaceae, Tovariaceae, Resedaceae, Gyrostemonaceae, Batidaceae, Moringaceae (position of the three last families somewhat uncertain but probably justifiable here; L. Bolt Jsrgensen, personal communication). SALVADORALES (position uncertain) : Salvadoraceae. THEIFLORAE THEALES : S tachyuraceae, Pentaphylacaceae (position uncertain), Marcgraviaceae, Quiinaceae, Ancistrocladaceae, Dioncophyllaceae, Nepenthaceae, Medusagynaceae (position uncertain), Caryocaraceae, Strasburgeriaceae, Ochnaceae, Oncothecaceae, Scytopetalaceae (position uncertain), Lecythidaceae (including Asteranthaceae, Foetidiaceae, Barringtoniaceae and Napoleonaceae), Theaceae (including Tetrameristaceae and Pellicieraceae), Hypericaceae (= Clusiaceae), Elatinaceae. DROSERALES: Droseraceae, Lepuropetalaceae, Parnassiaceae. PRIM ULI FLO RAE EBENALES: Ebenaceae, Sapotaceae, Styracaceae, Lissocarpaceae. PRIMULALES: Myrsinaceae, Aegicerataceae, Theophrastaceae, Primulaceae, Coridaceae. ROSIFLO RAE TROCHODENDRALES: Trochodendraceae, Tetracentraceae, Eupteleaceae, Cercidiphyllaceae. HAMAMELIDALES: Hamamelidaceae (including Rhodoleiaceae and Altingiaceae), Platanaceae, Myrothamnaceae, Geissolomataceae.

6 96 R. M. T. DAHLGREN FAGALES : Fagaceae, Corylaceae, Betulaceae. BALANOPALES : Balanopaceae. JUGLANDALES: Rhoipteleaceae, Juglandaceae. MYRICALES : Myricaceae. CASUARINALES : Casuarinaceae. BUXALES: Buxaceae, Daphniphyllaceae. CUNONIALES: Cunoniaceae, Baueraceae, Ribesiaceae, Brunelliaceae, Davidsoniaceae, Eucryphiaceae, Bruniaceae, Grubbiaceae. SAXIFRACILES: Crassulaceae. Cephalotaceae, I teaceae, Francoaceae, Saxifiagaceae, Vahliaceae, Greyiaceae. GUNNERALES: Gunneraceae. ROSALES: Crossosomataceae (position uncertain), Rosaceae, Neuradaceae, Malaceae (= Pomaceae), Amygdalaceae, Chrysobalanaceae. PODOSTEMIFLORAE PODOSTEMALES: Podostemaceae (including Tristichaceae). FABIFLORAE FABALES : Mimosaceae, Caesalpiniaceae, Fabaceae. PROTEI FLO RAE PROTEALES : Proteaceae. MY RTIFLO RAE MYRTALES: Myrtaceae (including Heteropyxidaceae), Psiloxylaceae, 0 liniaceae, Melastomataceae (including Memecylaceae), Penaeaceae, Crypteroniaceae, Lythraceae, Sonneratiaceae, Punicaceae, Combretaceae, Onagraceae, Trapaceae (position uncertain). HALORAGALES: Haloragaceae. RHIZOROPHORALES: Rhizophoraceae (excluding Anisophylloideae). RUTIFLORAE RUTALES : Rutaceae (including Rhabdodendraceae and Flindersiaceae), Cneoraceae, Surianaceae, Simaroubaceae, Burseraceae, Meliaceae (including Aitoniaceae). SAPINDALES: Coriariaceae, Anacardiaceae (including Pistaciaceae and Julianiaceae), Leitneriaceae, Podoaceae, Sapindaceae (including Stylobasidiaceae), Hippocastanaceae, Aceraceae, Akanaiceae, Bretschneideraceae (position uncertain), Emblingiaceae, Meliosmaceae, Staphyleaceae, Sabiaceae (position uncertain), Connaraceae, Melianthaceae (position uncertain). BALSAMINALES : Balsaminaceae. POLYGALALES: Malpighiaceae, Trigoniaceae, Vochysiaceae, Polygalaceae (including Xanthophyllaceae and Diclidantheraceae), Krameriaceae. GERANIALES: Zygophyllaceae, Nitrariaceae, Peganaceae, Balanitaceae, Erythroxylaceae, Humiriaceae, Linaceae, Ctenolophaceae, Ixonanthaceae, Lepidobotryaceae, Oxalidaceae (including Averrhoaceae), Geraniaceae, Dirachmaceae, Ledocarpaceae, Vivianiaceae, Biebersteiniaceae. TRO PAEOLALES (position uncertain) : Tropaeolaceae, Limnanthaceae. SANTALIFLORAE CELASTRALES : Celastraceae (including Hippoaateaceae, Tripterygiaceae, Siphonodontaceae and Goupiaceae), Stackhousiaceae, Lophopyxidaceae,

7 ANGIOSPERM CLASSIFICATION 97 Cardiopteridaceae (position uncertain), Corynocarpaceae (position uncertain). VITIDALES: Vitidaceae (including Leeaceae). SANTALALES : Olacaceae (including Octoknemataceae), Opiliaceae, Loranthaceae, Misodendraceae, Eremolepidaceae, Santalaceae, Viscaceae. BALANOPH ORIFLO RAE BALANOPHORALES (position uncertain) : Cynomoriaceae, Balanophoraceae. ARALIIFLORAE PITTOSPORALES: Pittosporaceae, Tremandraceae (position uncertain), Byblidaceae. ARALIALES : Helwingiaceae, Araliaceae, Apiaceae. ASTERIFLORAE CAMPANULALES: Pentaphragmataceae, Campanulaceae (including Sphenocleaceae), Lobeliaceae. ASTERALES : Asteraceae. SOLANIFLORAE SOLANALES: Solanaceae (including Nolanaceae, Duckeodendraceae), Sclerophylacaceae, Goetzeaceae (position uncertain), Convolvulaceae (including Humbertiaceae), Cuscutaceae, Cobaeaceae, Polemoniaceae. BORACINALES: Hydrophyllaceae, Ehretiaceae, Boraginaceae, Wellstediaceae, Lennoaceae (position uncertain), Hoplestigmataceae (position uncertain). CORNIFLORAE FOUQUIERIALES: Fouquieriaceae. ERICALES: Actinidiaceae (including Saurauiaceae), Clethraceae, Cyrillaceae, Ericaceae, Empetraceae, Monotropaceae, Pyrolaceae, Epacridaceae, Roridulaceae, Diapensiaceae. EUCOMMIALES : Eucommiaceae. SARRACENIALES: Sarraceniaceae. CORNALES : Garryaceae, Alangiaceae, Nyssaceae, Cornaceae, Aucubaceae, Corokiaceae, Davidiaceae, Torricelliaceae, Phellinaceae, Aquifoliaceae, Paracryphiaceae, Sphenostemonaceae, Symplocaceae, Anisophyllaceae (position uncertain), Icacinaceae, Escalloniaceae, Montiniaceae, Medusandraceae (position uncertain), Columelliaceae, Stylidiaceae (including Donatiaceae), Alseuosmiaceae, Hydrangeaceae, Dialypetalanthaceae, Sambucaceae, Adoxaceae, Dulongiaceae, Tribelaceae, Eremosynaceae, Pterostemonaceae, Tetracarpaeaceae (position and status of the last 5 families uncertain). LOASI FLO RAE LOASALES: Loasaceae. GENTIANIFLO RAE GOODENIALES: Goodeniaceae (including Brunoniaceae). OLEALES: Oleaceae. GENTIANALES: Loganiaceae (including Antoniaceae, Spigeliaceae, Strychnaceae, Potaliaceae), Rubiaceae (including Theligonaceae), Menyanthaceae, Gentianaceae, Apocynaceae, Asclepiadaceae.

8 98 R. M. T. DAHLGREN LAM I IFLO RAE SCROPHULARIALES: Bignoniaceae, Myoporaceae, Gesneriaceae, Buddlejaceae, Scrophulariaceae (including Nelsoniaceae and Orobanchaceae), Globulariaceae, Selaginaceae, Stilbaceae, Retziaceae, Plantaginaceae, Lentibulariaceae, Pedaliaceae, Trapellaceae, Martyniaceae, Acanthaceae, Thunbergiaceae, Mendonciaceae, Henriqueziaceae (position uncertain). HIPPURIDALES: Hippuridaceae. HYDROSTACHYALES (position uncertain, possibly allied to Cunoniales or Gunnerales in the Rosiflorae) : Hydrostachyaceae. LAMIALES: Verbenaceae (including Phrymaceae, but excluding Stilbaceae), Callitrichaceae, Lamiaceae. SUBCLASS LILIIDAE (=MONOCOTYLEDONEAEI ALI SMATIFLO RAE HYDROCHARITALES: Butomaceae, Aponogetonaceae, Hydrocharitaceae (incl. Thalassiaceae and Halophilaceae). ALISMATALES : Alismataceae (including Limnocharitaceae). ZOSTERALES: Scheuchzeriaceae, Juncaginaceae (including Lilaeaceae), Najadaceae, Potamogetonaceae (including Ruppiaceae), Zosteraceae, Posidoniaceae, Cymodoceaceae, Zannichelliaceae. TRIURIDIFLORAE TRIURIDALES: Triuridaceae. ARIFLORAE ARALES: Araceae, Lemnaceae. LILIIFLORAE DIOSCOREALES: Dioscoreaceae (including S tenomeridaceae), Trichopodaceae, Taccaceae, S temonaceae (including Croomiaceae), Trilliaceae. ASPARAGALES: Philesiaceae, Luzuriagaceae, Geitonoplesiaceae, Smilacaceae (including Ripogonaceae), Petermanniaceae, Convallariaceae, Asparagaceae, Ruscaceae, Herreriaceae, Dracaenaceae, Nolinaceae, Doryanthaceae, Dasypogonaceae, Xanthorrhoeaceae, Agadaceae, Hypoxidaceae, Tecophilaeaceae, Cyanastraceae, Phormiaceae, Dianellaceae, Eriospermaceae, Asteliaceae, Hanguanaceae (position uncertain), Aphyllanthaceae, Anthericaceae, Asphodelaceae (including Aloeaceae), Hemerocallidaceae, Funkiaceae, Hyacinthaceae, Alliaceae (including Agapanthaceae and Gilliesiaceae), Amaryllidaceae. HYDATELLALES (position uncertain) : Hydatellaceae. LILIALES: Colchicaceae, Iridaceae, Geosiridaceae, Calochortaceae, Tricyrtidaceae, Alstroemeriaceae, Liliaceae, Melanthiaceae. BURMANNIALES: Burmanniaceae, Thismiaceae, Corsiaceae. ORCHIDALES: Apostasiaceae, Cypripediaceae, Orchidaceae. VELLOZIALES : Velloziaceae. BROMELIALES: Bromeliaceae. HAEMODORALES: Haemodoraceae (including Conostylidaceae). PONTEDERIALES: Pontederiaceae. PHILYDRALES: Philydraceae. TYPHALES: Sparganiaceae, Typhaceae.

9 ANGIOSPERM CLASSIFICATION 99 ZING I B E RI FLO RAE ZINGIBERALES: LOwiaCeae, Musaceae, Heliconiaceae, Strelitziaceae, Zingiberaceae, Costaceae, Cannaceae, Marantaceae. COMMELINIFLORAE COMMELINALES : Mayacaceae, Commelinaceae (including Cartonemataceae). ERIOCAULALES: Rapateaceae, Xyridaceae, Eriocaulaceae. JUNCALES: Thurniaceae, Juncaceae. CYPERALES : Cyperaceae. POALES: Restionaceae (including Anarthriaceae and Ecdeiocoleaceae), Centrolepidaceae, Flagellariaceae, Joinvilleaceae, Poaceae. ARECIFLORAE ARECALES: Arecaceae. CYCLANTHALES: Cyclanthaceae. PANDANALES : Pandanaceae. The Magnolitjlorae SOME FEATURES OF THE CLASSIFICATION The Magnoliiflorae here comprise mainly the woody members of the so-called primitive dicotyledons. The following attributes are widespread, though lacking in various members: primitive wood; P-type plastids in the sieve elements; alternate exstipulate leaves ; cells with essential oils in leaves ; benzylisoquinoline alkaloids present; tepals occasionally with tendency to trimery ; stamens flattened and leaf-like with the microsporangia below the apex, pollen grains sulcate, 2- nucleate; gynoecium apocarpous or monocarpellate, endosperm formation cellular. Nelumbo has been attached here rather than in the Ranunculiflorae, with which its connection is probably not so strong; it is doubtless isolated and possibly has evolved its herbaceous habit independently from the Ranunculiflorae, even if its tricolpate pollen grains would be more in place in this superorder. The Nymphaezjlorae This comprises the Piperales and Nymphaeales, the former is undoubtedly also allied to the Magnoliiflorae. The most conspicuous feature of the Nymphaeiflorae is the voluminous starchy perisperm, on top of which there is still a well-defined endosperm tissue enclosing a small embryo. This represents an organization which is unlikely to have evolved by convergence in independent lines of evolution, especially when considering further concordant features in the two orders, such as the S-type sieve tube plastids, the sulcate (or zonisulculate) pollen grains and the similar leaves. Benzylisoquinoline alkaloids are rare in the Piperales (but known in a single species of Piper) and apparently totally absent in the Nymphaeales. In addition, leaves with essential oils are found in Piperales but not in Nymphaeales. However, the latter order has evolved the capacity to synthesize ellagitannins, otherwise unknown in this superorder, and absent in the Magnoliiflorae and Ranunculiflorae. Piperales and Nymphaeales are no doubt very different as a consequence of their adaptations to tropical rain forest and aquatic habitats respectively. While the former has experienced a strong floral reduction, some groups (especially Nymphaeaceae) have evolved large, diversified and polymerous flowers. A tendency for floral trimery is found in both orders.

10 100 R. M. T. DAHLGREN The Ranuncultjlorae This superorder has certain features in common with the Magnoliiflorae, such as the common presence of benzylisoquinoline alkaloids (here more complex) and the frequent apocarpy, trimery and spiral organization of floral parts. A few members have cellular endosperm formation. Few problems are involved in this superorder, although family rank is applied most inconsistently to generic groups in the Ranunculaceae-Berberidaceae region; families such as Ciraesteraceae, Kingdoniaceae, Glaucidiaceae, Hydrastidaceae, Podophyllaceae and Nandinaceae being acknowledge by some authors but not by others. The Papaverales are undisputably related to the Ranunculales. The Caryophylhj7orae This has been restricted to the order Caryophyllales, which is perhaps the best studied of all orders with regard to the families and their interrelationships (Mabry & Behnke, 1976). The order is easily defined by the characteristic P-type sieve element plastids in combination with the well-developed, starchy perisperm generally encircled by a curved embryo. Other common features are the presence of betalains and saponins, the commonly aberrent secondary growth and the frequently free, central placentation. By the exclusion of Batidaceae, Gyrostemonaceae, Theligonaceae, Polygonaceae, Plumbaginaceae, Fouquieriaceae, Frankeniaceae, Elatinaceae, etc., the circumscription is now clearcu t. The Polygoniflorae The superorder comprises the order Polygonales which has long been associated with Caryophyllales with which it shows several conspicuous similarities. Some of these are superficial, others evolved by convergence (such as the starchy seed) and some may express phylogenetic relationships. To the latter category belong the amino acid sequence (Boulter, 1973) as well as shared parasitic fungi (e.g., aecidia of grass rusts). The Malvtjlorae The Malviflorae (=Dilleniiflorae) are here very broadly conceived, and the nine orders show affinity in different features. The order Cistales has here been dissolved, since the Cistaceae show great resemblance in morphological, anatomical, embryological, and chemical characters (also in flavonoids, cf: Gornall, Bohm 8c Dahlgren, 1979) with families of the Malvales. They approach closely the Bixaceae as well as the Cochlospermaceae, Tiliaceae and Sterculiaceae. The Dilleniaceae are less clearly allied to the Malvales and form a separate order, but agree to some extent with malvalean taxa so that a position in this vicinity is justified. The genus Paeoniu, making up the Paeoniales, is placed with some reservation in conjunction with Dilleniales. The close affinity between the Malvales and Euphorbiales and the affinity between these and the Urticales (Berg, and Thymelaeales (Hegnauer, 197.3) is now becoming increasingly clear. The order Rhamnales is here restricted to include Rhamnaceae only. It was acknowledged to be related to Euphorbiales by Thorne ( 1976) by virtue of several features, among which should be mentioned the mucilage cells, the phloem, the lepidote vesture, and some unusual peptide alkaloids. The Elaeagnales are no

11 ANGIOSPERM CLASSIFICATION I 0 I doubt best treated in, or next to, this order, but the Plumbaginales have a more uncertain position. The embryology including solitary anatropous ovules, instability in embryo sac formation, and presence of an obturator may be taken to support a euphorbiaceous affinity, while its seed coat structure is more similar to that in the Rhamnaceae or Elaeagnaceae. The Viollflorae The Violiflorae consist of various orders with 2- or 3-carpellate gynoecia having parietal placentation. A tendency for producing cyanogenic glucosides or/and glucosinolates is apparent in some orders. Gynophores and androgynophores are common. Ellagitannins are usually absent (except in the Tamaricales), The position of Gyrostemonaceae and Batidaceae (two small glucosinolate families mainly in the Southern Hemisphere) in Capparales is still questionable but a certain affinity with Resedaceae seems. supported (Bolt Jsrgensen, personal communication). The Thelflorae The Theales form a fairly homogeneous group of mostly woody plants, frequently with scalariform perforation plates in the vessels, simple, entire and coriaceous leaves, and choripetalous flowers with secondarily multistaminate androecia. The ovules are mostly bitegmic and tenuinucellate, a fairly unusual combination otherwise found mainly in the Primuliflorae among dicotyledons. The Primullflorae This group, consisting of the probably closely allied Ebenales and Primulales, generally has actinomorphic and sympetalous flowers with central, sometimes free placentation. The ovules are bitegmic and mostly tenuinucellate. The close relationships of this group to other sympetalous groups have been much debated but are dubious, as there are weak embryological and chemical bases for an alignment to most of these groups. The chemical characters are more like those of the Theales. The Rosgorae This superorder has been extended here to include the equivalent of the Hamamelidiflorae and Saxifragiflorae (of Dahlgren, 197 5a). These complexes are obviously allied through series of families. The trochodendralean end of this complex by its primitive xylem, its cellular endosperm formation and apocarpy are frequently associated with the magnoliiflorean complex (Endress, 1969) but deviate from them in pollen morphology, chemistry and other details. There is still some doubt about their mutual relationships and also about their affinity to the Hamamelidales. The Hamamelidales combine features of the mainly insectpollinated Cunoniales-Rosales lines with the wind-pollinated Fagales, with which they are undoubtedly related (Endress, 1977). The position of the Myricales and Juglandales, which agree in many respects, has been interpreted differently, mostly in favour of a fagalean but also of a sapindalean (-anacardiaceous) relationship (Thorne, 1974). Recent serological indications have been decisive in placing them next to Fagales (Fairbrothers and Peterson, personal communications). Balanopales (= Balunops) is at present under in-

12 102 R. M. T. DAHLGREN vestigation (Carlquist, personal communication). It has some features of the Fagales, others of the Cunoniales, and might be best placed between these orders. The Podostemonzforae Podostemonales (= Podostemonaceae sensu lato) are so specialized that they cannot be associated with any other superorder without severe reservations. The specialized habit and lack of endosperm formation are outstanding features. These, in combination with the silica bodies, the frequently trimerous gynoecium and fusion of stamen filaments could lead the imagination towards the orchids, with which they have virtually no affinity (the seedlings for example being typically dicotyledonous). The FabiJorae Although the order Fabales is often associated with the Rosales, its connection with the Sapindales also seems to gain an increasing support through morphological, palynological, anatomical, embryological and chemical evidence. Thus the argument as to whether the Connaraceae should be placed with the Fabales or with the Sapindales becomes less controversial when the orders are now acknowledged as closely allied (and placed in close conjunction in a diagram). The Proteijorae There is little doubt that the disjunct Proteaceae is an old, isolated taxon. It is therefore placed in a separate order, Proteales, in its own superorder. As in the Fabales, the pistil is monocarpellate and its solitary to several ovules have a similar embryology. The chemical attributes, including flavonol derivatives, leucoanthocyanins, arbutin and tannin accumulation agree better with a position near the Fabales and the Rosales than in conjunction with the Santalales or the Thymelaeales with which the Proteales have sometimes been associated. The MyrtzJlorae The Myrtales comprise a fairly natural group or perhaps two related groups of families (Briggs & Johnson, 1979) which frequently possess opposite leaves, rudimentary stipules, intraxylary phloem, tetramerous flowers with a hypanthium, and two stamina1 whorls. The chemistry is related to that of the Rosales, with which the Myrtales are perhaps the most closely allied. The Haloragaceae and Rhizophoraceae senm strict0 deviate from the myrtalean families to a variable extent and have a more uncertain position. However, their position next to the Myrtales is likely, which explains their treatment here as separate orders in this superorder. The Rutzjlorae This complex is one of the most variable in the angiosperms. In spite of this, families in the six orders approach each other to such an extent that it is difficult to draw clear borderlines. Thus, the Zygophyllaceae are sometimes placed in the order Rutales, sometimes in Geraniales ; the Anacardiaceae are sometimes placed in Rutales, sometimes in Sapindales. The flowers which are mostly choripetalous, pentamerous and frequently obdiplostemonous and the frequent occurrence of

13 ANGIOSPERM CLASSIFICATION 103 aromatic or resinous compounds are fairly typical. The two families of the Tropaeolales are characterized by glucosinolates. The superficial similarity to Balsaminaceae may or may not indicate relationship to this familv, and their affinity to the Geraniales is also dubious. The Santaltjorae The order Celastrales in the present classification has been liberated from a number of small families of uncertain affinity included there in Dahlgren (1975a1, and now consists chiefly of the Celastraceae. Several attributes indicate that this may be allied to the Sapindales (where in the future it may prove to be more properly placed). Its long-acknowledged similarities to Vitidaceae and the less advanced families of Celastrales (Olacaceae, Opiliaceae) justify the present superorder. The Vitidaceae, by their combination of common attributes, are difficult to place, and the problem has not decreased since Behnke (1974) discovered that the sieve tube plastids are of the P-type. The Santalales no doubt comprise a homogeneous group of families showing a successive reduction of ovules, culminating in the Viscaceae and Loranthaceae. Polyacetylenes and certain alkaloids also connect the santalalean families. The Balanophoriflorae As is the case among the semiparasitic taxa of the Santalales the wholly parasitic Balanophorales (Balanophoriflorae) have undergone a strong ovular reduction. Whereas the ovules of Cynomoriaceae are still unitegmic those of Balanophoraceae are constantly ategmic and normally not well differentiated from the inner tissue of the ovary, the climax being the extremelv small, archegonium-like female flowers in Balanophora (Hansen, 1972). The affi nities of the Balanophorales are still totally unknown (as for the Podostemonales above). Some have claimed that the Balanophoraceae and Cynomoriaceae are wholly parasitic members of the Santalales, with which there are embryological resemblances; other possible relatives are the Gunneraceae, or possibly the Ericales or even the Asterales. The AralizfEorae By the common presence of resinous ducts, polyacetylenes and a similar basic plan in the flower construction, this group is likely to form an acceptable unit (Hegnauer, 1969), but the circumscription of Pittosporales present some problems. The core group of the Araliiflorae is the Araliales, with the closely allied Araliaceae and Apiaceae recently united by Thorne (1973) to a single family. A relationship of the Araliales to the Asterales is likely from their similar basic floral construction, similar embryological features, 3-nucleate pollen grains, exstipulate leaves, multilacunar leaf tr aces, and possession of polyacetylenes and sesquiterpene lactones. The AsterzfEorae On the whole the basic construction of the flowers in the Asterales and Campanulales is similar and the orders exhibit parallel cases of pollen pump mechanisms, parallel types of floral zygomorphy, etc. The chemical features are variable in both orders but exhibit some similarities such as the ability to

14 104 R. M. T. DAHLGREN form polyacetylenes but not iridoids. It should be noted that the Campanulales as circumscribed here do not include Calyceraceae, Stylidiaceae or Goodeniaceae, three families which have often, no doubt incorrectly, been placed in Campanulales on the basis of superficial resemblance. They are now placed in different positions in the Gentianiflorae-Corniflorae complexes, where their inclusion involves no strain. As mentioned above, the Asterales also show very strong similarities to the Araliales. The SolanaJorae The Solanales and Boraginales could probably be treated in the same order (cf: Dahlgren, 1975a), although a subdivision is practical and easy, if based on the ovules, which have their micropyle directed downwards (Solanales) and upwards (Boraginales) respectively. The sympetaly and the unitegmic, usually tenuinucellate ovules with cellular or nuclear endosperm formation are strongly reminiscent of those in the Asteriflorae as well as those in the Lamiiflorae or Gentianiflorae, but as far as is known, no taxa of the Solaniflorae possess the polyacetylenes (or sequiterpene lactones) of the Asteriflorae nor the iridoids of the Lamiiflorae or Gentianiflorae. Instead they produce a multitude of other poisonous substances. The CornEflorae This complex is a variable one with unitegmic tenuinucellate ovules, cellular endosperm formation, frequent sympetaly and possession of endosperm haustoria as the most characteristic attributes. Many taxa form iridoid compounds, which are exclusively of carbocyclic types, except in the Sarraceniales, some Cornales and most Dipsacales, where secoiridoids are formed. Most taxa of the Corniflorae are woody. The vessels, especially in Ericales and Sarraceniales, may be primitive, with oblique end walls having scalariform perforation plates with numerous bars. The seeds may either be small (many Ericales) or may have a small embryo in combination with copious endosperm, although in some Cornales and Dipsacales there is a tendency for the endosperm to be consumed and for the embryo to increase in relative size. The flowers also show a great variation. They are mostly pentamerous and haplo- to obdiplostemonous, but a multiplication in the androecium has occurred in some families (e.g., Actinidiaceae and Hydrangeaceae). The Fouquieriales in spite of their probably bitegmic ovules have iridoids, ellagic acid, tricarpellate pistil, and 10 or more stamens and other attributes in which they agree with the Cornales- Ericales (Dahlgren et al., 19761, but not with the Solaniflorae next to which they are sometimes placed. The Loacijlorae The Loasaceae in most classifications are placed in the Violales (Parietales), where they are divergent especially in their embryologic features, the ovules being unitegmic and tenuinucellate, the endosperm formation cellular, and terminal endosperm haustoria present. Takhtajan ( 1969) placed the family near Boraginaceae, which is more, but not wholly, satisfactory from this point of view. The discovery of iridoids in Loasaceae points even more strongly to a position in the region of the Corniflorae-Lamiflorae-Gentianiflorae, and also the recent

15 ANGIOSPERM CLASSIFICATION discovery of secoiridoids, indicates a Cornalean-Dipsacalean or Gentianiflorean affinity. Several genera of Loasaceae have haplostemonous androecia, which is in better accord with the condition in the mentioned orders, although these have sympetalous flowers whereas those in Loasaceae are choripetalalous. This fact, in combination with a few attributes often found in the Violales, has been decisive in placing the Loasaceae in a separate order and superorder. The Gentiany7orae The main part of the Gentianiflorae is made up of the Gentianales, which include among other families the Rubiaceae. Some interesting though not consistent attributes in this order are the opposite leaves, (interpetiolar) stipules, intraxylary phloem, secoiridoids and, more rarely, indol and ipecac alkaloids. The endosperm formation is also mostly nuclear in the Gentianales while it is cellular in Oleales, Goodeniales and in the other superorders containing iridoids. The orders of the Gentianiflorae show important similarities in many features and contain secoiridoid producing plants. Thus the Oleaceae, for example, are very similar to the Loganiaceae in many respects. The Lamiy7orae In contrast to the Gentianiflorae, the Lamiiflorae do not form secoiridoids but only carbocyclic iridoids, which are somewhat irregularly scattered through the superorder. The flowers tend to be zygomorphic and are aggregated in racemose inflorescences (thyrses, racemes, spikes, heads). As in the previous orders the ovules are unitegmic, tenuinucellate and have cellular (rarely intermediate) endosperm formation. In the Lamiales the fruit tends to be. a four-seeded schizocarp as in (but independently ofl the Boraginales of the Solaniflorae. The Hippuridales, and even more so the Hydrostachyales, are placed in Lamiiflorae with reservations. The former is perhaps related to Plantaginaceae, and the Hydrostachyaceae were also considered to be allied to this complex by Jager- Zurn (1965), but other conclusions based on the same data (see, for example, Thorne, 1968) are that the Hydrostachyaceae approach Myrothamnaceae and allied families in the Rosiflorae. Perhaps Gunneraceae should also be considered as possible relatives of the Hydrostachyales. The AlismatGorae This superorder is easily circumscribed. It is characterized by a number of attributes which are otherwise unusual in monocotyledons : presence of intravaginal squamules; absence of vessels in the stems (and often in the roots as well); root hairs which are attached to particular, short epidermis cells; amoeboid tapetum ; 3-nucleate pollen grains; apocarpy or monocarpellate gynoecia ; caryophyllad type of embryo formation, and absence of oxalate raphides and silica bodies, etc. The Alismataceae are distinct from the other families in several attributes, and make up the Alismatales, while the Zosterales are more gradually attached to some Hydrocharitales. The Alismatiflorae are probably most closely related to the Ariflorae. The TriuridiJEorae There is little doubt that the Triuridales are at least distantly related to the Alismatiflorae but the closeness of the relationship is uncertain. The superorder I05

16 106 R. M. T. DAHLGREN consists of chlorophyll-less saprophytes growing in tropical rainforests. The flowers are small and unisexual, the female ones having numerous monocarpellate, often subgynobasic pistils. This group differs from the Alismatiflorae in the endospermous seeds; absence of parietal cells, and periclinal divisions in the epidermis of the nucellus. The ArtJlorae Consisting of the Arales (Araceae and Lemnaceae) only, this is a well-defined complex. It is suprisingly similar to the Alismatiflorae in a number of concealed characters recalling the kind of similarity between the Piperales and Nymphaeales. Such attributes are the vessel-less stem, amoeboid tapetum, and the caryophyllad type of embryogeny. Typical of the Arales are the spadix supported by a spathe; baccate fruits; tetracytic stomata; cellular endosperm formation, and copious amounts of oxalate raphides; in all of which they are different from the Alismatiflorae. The Lilitjorae The Liliiflorae comprise a considerable part of the monocotyledons including for example the huge order Orchidales. Generally the stems are vessel-less ; buibs, corms and rhizomes are common, though in different groups. Normally (but not consistently) the flowers have two petaline tepal whorls, which may or may not be similar in shape and size. Septa1 nectaries or nectaries at the tepal base are found in diverse groups, and the pollen grains are more often than not monosulcate. The stomata are usually anomocytic, and certain orders have a starchy endosperm (Bromeliales, Haemodorales, Velloziales, Pontederiales, Philydrales and Typhales). In other classifications these are often associated with the Commeliniflorae, and are transitional between for example the Commelinales in that superorder and the Asparagales and Liliales of the Liliiflorae. The Zingiberqorae This superorder is well-defined by the combination of arillate and frequently perispermous seeds, petiolate leaves, and a tendency for some of the stamen homologues to be petaloid. The members with 5-6 functional stamens have oxalate raphides and their seeds lack perisperm but have copious endosperm. In these and many other attributes these families approach some liliiflorean orders (Pontederiales, Bromeliales, Philydrales). The CommelintJlorae In contrast to the Liliiflorae, this superorder shows adaptations to wind pollination, and even those groups which are insect-pollinated, such as the Commelinales, are difterent from the Liliiflorae, having, for example, sepaline outer tepals. Most Commeliniflorae differ from the majority of the Liliiflorae in characters such as the common occurrence of silica bodies, absence of oxalate raphides, presence in stems of vessels with simple (and scalariform) perforation plates, paracytic stomata, ulcerate pollen grains, starchy endosperm, absence of steroid saponins and absence of chelidonic acid. Within the Commeliniflorae, the Juncales-Cyperales, the Commelinales- Eriocaulales and the Poales (including the Restionales) seem to form three fairly well-defined subgroups.

17 The Areclflorae ANGIOSPERM CLASSIFICATION 107 This last superorder is retained in its classical strict circumscription, i.e., with the exclusion of the Typhales and the Arales. The group consists of large herbs to little-branched trees, the stems of which have no secondary thickening growth, cf some woody members of the Asparagales (of the Liliiflorae). The stomata are largely tetracytic, silica bodies are frequently present, septa1 nectaries are common (insect pollination is fairly wide-spread), the pollen grains are 3- nucleate, the locules one-ovuled, the fruits drupaceous or baccate and the endosperm generally non-starchy. THE. DIAGRAM, POSSIBILITIES AND LIMITATIONS In the diagram of Fig. 1 the orders are illustrated as the transections of branches of an imaginary evolutionary tree or hedge. The position of the orders in relation to each other has been determined as far as possible by the degree of similarity, reflecting the supposed phylogenetic relationships. In addition, the depicted size of the orders is, to a reasonable degree, proportional to the number of species included, but not strictly, because the area of the orders with one or few species would then be too small to permit any shading or symbol to be plotted in it. As mentioned earlier (Dahlgren, 1975a: 122) such a framework in two dimensions cannot possibly fulfil all requirements of juxaposition. An order, when placed appropriately in relation to two or three orders, is often not allowed the desired proximity to a fourth order according to its similarities to other orders. The juxapositions become self-explanatory when a sufficient number of phylogenetically significant character states are followed in the diagram, and this is why the distribution of some characters have been included (Figs 3-7). Despite the limitations of the diagram, it is useful in the following respects: (1) It permits a quick and effective survey of the distribution of any mapped attribute. (2) It permits effective comparisons between the distribution of two, or several, such attributes, whether these show great coincidence, are wholly or partly exclusive, or, as in many cases, show other patterns. (3) It offers a chance to predict a yet-unknown character state in a species ofa family from known characters in the neighbouring families, provided that there is a clear tendency for aparticular state of this character to dominate in the order (superorder). (4) The diagram, when used for mapping character states considered as typical to an order, may be used to test homogeneity. Each family consistently has a definite area and position within the ordinal figure. Where the same family deviates in several features from the other families this can quickly be recognized as a spot of deviant shading in one diagram after another. Its position then must be reconsidered. (5) As a result of comparison ofnumerous such diagrams, a deeper insight into relationships between larger groups (orders, superorders) may be possible; this diagram then serves as a kind of punch-card system with a common matrix. 8

18 Figure 1. The superorders of the angiosperms illustrated as clusters of orders in a two-dimensional frame-work. This diagram should be complemented with the following diagrams in which the order names are given and with the linear list of orders and families in the text.

19 ANGIOSPERM CLASSIFICATION I09 One of the limitations with the framework is that it is tempting to memorise it as a factual constellation, while in fact, it serves as only one of several possible ways of illustrating the ordinal relationships. Though superficially very different from my diagram of 1975 (Dahlgren, 1975a1, this diagram depends on much the same base material, and most of the main orders and superorders cohere in a similar way. Both diagrams thus serve the same purpose although the latest version is superior in many respects. Many requirements must be laid on a diagram of this sort in order to make maximum use of it. There is obviously an upper tolerance limit for the unnaturalness of its orders. For example, should the families of Scrophulariales be intermingled with those of Gentianales, then characters like the pattern of endosperm formation, types of iridoids, floral symmetry, etc. would give confusing patterns - although by correlating the patterns of these characters it might be possible to reconstruct the orders as circumscribed here. Similar two-dimensional diagrams (Stebbins, 1974; Thorne, 1976, 1978, 1979, in press) are capable of serving the same purpose. The most severe limitation of the diagram is that false conclusions may be drawn from individual features, or positions of orders in it. As was mentioned above, for spatial reasons some orders come in closer conjunction with each other than would be justified from an all-round comparison between them, others are more distant than isjustified. This difficulty could be solved partly by extending the figures for the orders as was done in Dahlgren, 1975a, for Saxifragales, Geratiales, Sapindales and Hamamelidales, but this method has been avoided here, because the false conclusions of linear sequence were often drawn for the families of these orders. Complex patterns are difficult to illustrate with this, or any other, method. DISTRIBL TIONS OF SOME ATTRIBUTES IN THE DIAGRAM Only the distributions of a few attributes can be shown in this presentation, and therefore some of the phylogenetically more significant ones have been chosen, namely sympetaly (Fig. 21, unitegmic and ategmic ovules (Fig. 3), the occurrence of a parietal cell (Fig. 41, cellular and helobial endosperm formation (Fig. 5) and some chemical characters : iridoids, polyacetylenes, glucosinolates and benzylisoquinoline alkaloids (Fig. 6). In addition, for the monocotyledons, further character states will be shown (Fig. 7). As the distributions of these characters have been commented on elsewhere (Dahlgren, 1975b; Nielsen et al., 1975; Dahlgren, 1977a, b) only briefcomments are given here. Sympetalous flowers Syrnpetalous flowers have long been recognized as significant in classifications ever since pre-linnean times. Most taxa with sympetalous flowers have been classified in the Monopetalae, Metachlamydeae, Sympetalae in various of the classical works, and they have been grouped in the subclass Asteridae by Cronquist (1968) and Takhtajan (1969). Because of the correlation in a great many families between sympetaly, flowers with 5 sepals, 5 petals and 5 stamens in alternating whorls, and a bicarpellate pistil, these families have been recognized as a more or less uniform group. This impression has been supported even more

20 Figure 2. Approximate distribution of sympetalous flowers in the dicotyledons (the monocotyledons are excluded in this survey as any comparison would be irrelevant). Fusion of only some of the petals, as in Baslsaminaceae, Fabaceae smru stricto. Polygalaceae etc., is excluded. Problems arise where there are doubt of homology of the perianth, as in the Santalalg where Loranthaceae is shaded but not Viscaceae. The perianths ofthe Thymelaeaceae and the Proteaceae are interpreted as homologous with the calyx in other plants.

21 Figure 3. Approximate distribution of unite~pic (dots) and atepic (hatching) ovules. The great homogeneity of unite<pir ovuled groups in the right part of the diagram (consisting of mainly the sympetalous groups) is obvious. Unitegmy has arisen along various other lines of evolution and has evolved in different W~VP

22 112 R. M. T. DAHLGREN strongly by the concomittant possession of unitegmic, mostly tenuinucellate, ovules and a generally cellular type of endosperm formation. As will be shown below, some of the sympetalous groups are undoubtedly or probably unrelated to the main part of these Sympetalae, for example, the embryologically very aberrant Plumbaginales; the mostly bitegmic Primuliflorae, which, in addition, normally have nuclear endosperm formation; certain Mimosaceae (Fabales); The Cucurbitaceae and the Caricaceae (Violales), and many Crassulaceae (Saxifragales). The sympetalous monocotyledons are not included in the diagram of Fig. 3. Unitegmic ovules Figure 4 shows the distribution of ovules with one integument, a supposedly derived character state which no doubt has evolved along several evolutionary lines, one (or possibly more) of which has grown to be large and diversified. This coincides mainly with the Sympetalae complex, but also includes, for example, some choripetalous taxa included here with some sympetalous ones in the Corniflorae (comprising Ericales, Sarraceniales, Cornales, Dipsacales and Eucommiales). As will be shown below, the close mutual connection between the families included in Corniflorae and the relationships between these and the main part of the Sympetalae is supported by other embryological properties ( Philipson, 1974). Besides, the relationships between them and the Loasiflorae, Gentianiflorae and Lamiiflorae is supported by their contents of iridoids (see below). However, unitegmic ovules have also evolved in widely different and independent lines of evolution, among which should be mentioned many Fagales and all Juglandales and Myricales (the latter two orders being undoubtedly closely interrelated), also certain genera of Rosaceae, certain genera of Ranunculaceae (Bauman 8c Calis, , Pepermia and related genera ( Perperomiaceae ) of Piperaceae, most taxa of Salicaceae, the Sapotaceae, etc. Isolated genera with unitegmic ovules are further known in several other families. Possibly related to the Sympetalae are such groups as the Santalales and/or Balanophorales where the ovules are subjected to more or less strong reductions. Presence of parietal cells The dicotyledonous groups without parietal cells ( tenuinucellate ovules in the sense of Dahlgren, 1975b) again include most of the Sympetalae, including the ericalean and most cornalean families. Further, some dubiously related sympetalous groups lack parietal cells, such as most Primuluflorae. Besides, parietal cells are lacking in groups probably very distantly related to the Sympetalae, including most Theales and Droserales ; the Podostemonales ; some Geraniales (Oxalidaceae, Linaceae), all Tropaeolales and Balsamhales, and certain Celastrales. In monocotyledons, a parietal cell is lacking in the ovules of all the Orchidales, Burmanniales and Triuridales. It is also absent in the Poales and many Eriocaulales, Liliales, Asparagales and Velloziales. Endosperm formation Adding to the phylogenetic significance of the correlation between sympetaly and the two embryological characters previously mentioned is the distribution of dif ferent types of endosperm formation (Fig. 5). The ab initio cellular type is

23 Figure 4. Approximate distribution ofovules with and without a parietal cell in the angiosperms (unshaded and shaded respectively). Explanation is given in the text. Absence 01 parietal cells in dicotyledons partly coincides with the occurrence of one integument and is no doubt a derived character state. Groups with this attribute doubtless have evolved along various lines of evolution, one or a few main ones coinciding with the main sympetalous section. - w

24 Figure 5. Approximate distribution of the cellular (dots), helobial (vertical hatching) and intermediate (oblique hatching) types of endosperm formation in angiosperms. Unshaded areas represent the nuclear type plus unknown groups. Here the cellular type, represented in most Magnoliiflorae may well be the primitive state. Again, we find a concentration of one character state (the cellular type) in the Sympetalae, although, characteristically, it has broken down partly in the Asterales, Solanales and Boraginales and almost completely in the Gentianales.

25 ANGIOCPEKM CLASSIFICATION 115 concentrated in the Sympetalae, especially in the four iridoid-containing superorders Corniflorae, Lamiiflorae, Loasiflorae and Gentianiflorae (excepting the Gentianales in which endosperm formation largely has gone over to the ab initio nuclear type). The variability in endosperm formation in the Asterales of the Asteriflorae and the Solaniflorae is noteworthy, but the original type is probably the cellular type in these groups too. In addition, this type is found in Balanophorales, most Santalales, some Celastrales, and scattered members of the Rosiflorae sensu lato, the latter including Gunnerales, many Saxifragales, many Buxales, some Hamamelidales and the Trochodendrales. Further, in the dicotyledons, the cellular endosperm formation is found in many Magnoliiflorae and Nymphaeiflorae and some Ranunculiflorae. This pattern of distribution is doubtless of great phylogenetic interest in regard to the Sympetalae and related groups. On the basis of the common presence of cellular endosperm formation in these undoubtedly advanced plants, this character state is usually considered to be derived, but this is contradicted by the fact that the cellular type of endosperm formation is widespread in the Magnoliiflorae and Nymphaeiflorae. The fact that it is also found in the Trochodendrales, which are primitive in several respects, likewise supports the impression that the cellular type can be interpreted as primitive, and may have been retained in some evolutionary lines including the Sympetalae and others mentioned. In monocotyledons, the ab initio cellular endosperm is largely restricted to the Arales, but it is uncertain whether it is primitive here or not. The helobial endosperm formation is mainly restricted to the monocotyledons, and has its main distribution in the Alismatiflorae, the capsule-fruited Asparagales, the liliiflorous orders with starchy endosperm, the Zingiberales, the Juncales and the Cyclanthales. Most other monocotyledons have nuclear endosperm formation. Some chemical compounds Chemical characteristics can further elucidate the relationships, especially between the sympetalous taxa of the dicotyledons. IRIDOIDS comprise a group of chemical compounds which frequently serve as bitter principles in plants, making these unpleasant to browsing and egg-laying animals. Their distribution is shown in Fig. 6. While they are widely distributed in the Corniflorae, Loasiflorae, Lamiiflorae and Gentianiflorae, they are totally unknown in the Araliiflorae, Asteriflorae and Solaniflorae. Further, with the exception of four known genera (Liquidambar, Daphniphyllum, Xylocarpus, Fouquieria) they are entirely restricted to taxa with one integument. Their widespread distribution among the families in the four superorders suggests that they may have evolved early in their common ancestry. It is as yet difficult to interpret the odd occurrence of iridoids in the four genera outside these main groups; independent evolution in all or in some of them must not be excluded. POLYACETYLENES (Fig. 7) are largely vicarious in relation to the iridoids in the Sympetalae and their relatives, being found in the Asteriflorae, the probably quite closely related Araliiflorae, and in the Santaliflorae; otherwise, they are known only in very scattered genera (within the Rutales, Malvales, Laurales and Goodeniales). The discovery of polyacetylenes in the Simaroubaceae of the Rutales may have some significance.

26 Figure 6. Approximate distribution of certain groups of chemical compounds in angiosperms: iridoids (sparse dots), polyacetylenes (oblique hatching), glucosinolates (vertical hatching), and benzylisoquinoline alkaloids (dense dots). Comments are given in the text.

27 ,I 'VGI OSPERM CLASSIFICATION Figurc 7. Appr-oximdtc distribution of A, oxalate raphides and B, silica bodies in thr monocotvledons. The peculiar distribution of iridoids and of polyacetylenes may suggest either a possible bifurcation in the past evolution of the pro-sympetalae or a biphyletic origin of them. The Solaniflorae, which lack iridoids as well as polyacetylenes, may represent a separate evolutionary branch, or may have been connected to

28 118 R. M. T. DAHLGREN either of the other main groups, eventually having lost the ability to synthesize the characteristic compounds. The GLUCOSINOLATE or isothiocyanate plants include members of the orders Capparales (Brassicaceae, Capparaceae, Tovariaceae, Resedaceae, Moringaceae, Gyrostemonaceae and Batidaceae), Salvadorales (Salvadoraceae), Violales (Caricaceae), Euphorbiales (Drypetes of the Euphorbiaceae), Tropaeolales (Tropaeolaceae and Limnanthaceae) and perhaps Sapindales (Bretschneideraceae). There are still great doubts about the significance of glucosinolates. They comprise very diverse groups of compounds and may have evolved along several lines of evolution. Finally, the BENZYLISOQUINOLINE ALKALOIDS should be mentioned. They have their greatest concentration in the Magnoliiflorae and Ranunculiflorae, which are now generally believed to be related. However, they may be lacking in some magnoliiflorean families, such as Myristicaceae, Winteraceae, Illiciaceae and Schisandraceae, and several small families, not yet investigated. Further, they are unknown in all Nymphaeales and most Piperales, being so far known only in some species of Piper. In the Ranunculiflorae the benzylisoquinoline alkaloids are more complex than in the Magnoliiflorae, especially so in the Papaverales; but they are absent in part of Ranunculaceae and Berberidaceae (e.g., in Podophyllum) and not known in the Sargentodoxaceae or Lardizabalaceae, although they are plentiful in the Menispermaceae. The occurrence of complex benzylisoquinoline alkaloids in a few Rutaceae has been regarded as expressing relationships with the Ranunculiflorae, while their occurrence in other groups, (e.g., in some Rhamnaceae, members of the Buxaceae and Symplocaceae, Croton in the Euphorbiaceae) has not been considered to have any phylogenetic meaning. THE MONOCOTYLEDONS In the course of 1978 and 1979 the distribution of numerous attributes in the monocotoyledons has been studied more intensively in Copenhagen (Dahlgren & Clifford, in press). In the angiosperm classification presented here, the monocotyledons have been classified on the basis of these results. A separate diagram for monocotyledons has been used in Fig. 7 to illustrate the distribution of a few attributes in these. It seems that the classification of the monocotyledons does not present such great difficulties as that of the dicotyledons, although there are disagreements in finding an appropriate level when handling the Liliiflorae families. In the classificatory list the aim has been to circumscribe the families as naturally and homogeneously as possible. The opposite attitude would be to include most monocotyledons with two hypogynous petaline perianth whorls and two whorls of stamens in a giant Liliaceae smu lato. Such a concept would indeed represent a heterogeneous group and its maintenance deserves little support. An intermediary approach might cluster together the segregate groups of the Liliiflorae into somewhat larger families as follows : DIOSCOREALES: Dioscoreaceae (including Stenomeridaceae and Trichopodaceae), Taccaceae, Stemonaceae (including Croomiaceae), Trilliaceae ASPARACALES: Philesiaceae (including Luzuriagaceae and Geitonoplesiaceae), Smilacaceae (including R'ipogonaceae and Petermanniaceae), Convallariaceae

29 /i 1\1 L I 0s PERM CLASS1 FICATIO N I19 (including Asparagaceae, Ruscaceae and Hereriaceae), Dracaenaceae (including Nolinaceae), Doryanthaceae, Dasypogonaceae, Xanthorrhoeaceae, Agavaceae, Hypoxidaceae, Tecophilaeaceae, Cyanastraceae, Asphodelaceae (priority dubious; including Phomiaceae, Dianellaceae, Eriospermataceae, Asteliaceae, Aphyllanthaceae, Anthericaceae and Aloeaceae), Hanguanaceae, Hemerocallidaceae, Funkiaceae, Hyacinthaceae, (including Agapanthaceae and Gilliesiaceae), Amaryllidaceae LILIALES: Iridaceae, Geosiridaceae, Alstroemeriaceae, Colchicaceae, Liliaceae (including Tricyrtidaceae and Calochortaceae), Melanthiaceae BURMANNIALES : Burmanniaceae (including Thismiaceae), Corsiaceae ORCHIDALES: Apostasiaceae, Orchidaceae (including Cypripediaceae) TYPHALES: Typhaceae (including Sparganiaceae) (0 ther liliflorous orders are monofamilial.) This classification, however tempting, camouflages the great distinctness of many groups of genera (especially those placed here in Asphodelaceae ~enm lato) and even this semi-lumping approach is therefore discouraged here, as it would also be by, for example, Huber (1969, 19771, who has spent much effort in the study of the Liliiflorae. Many attributes have a phylogenetically significant distribution in the monocotyledons and may be useful in the main grouping. Among these only two will be chosen here, namely the distributions of oxalate raphides and of silica bodies. These both contribute in demonstrating a basic division between the commeliniflorean and the liliiflorean complexes (to the left and right respectively in the monocotyledon diagram). OXALATE RAPHIDES are wide-spread in the Liliiflorae but seem to be absent in most or all Commeliniflorae as here defined. It is noteworthy that raphides are also present in the Areciflorae and Ariflorae but lacking in the Alismatiflorae and Triuridiflorae complexes, while in the Zingiberiflorae they are restricted to the taxa with 5-6 anthers. SILICA BODIES show a more complicated distribution pattern, being mainly concentrated in the Commeliniflorae, the Zingiberiflorae, and at least the Arecales of the Areciflorae. However, there are silica bodies in the Bromeliales and many tropical orchids (Orchidales) among the Liliiflorae as circumscribed here. In order to evaluate this feature, the shape and size of the silica bodies, and also the position in the tissues of the cells containing them must be taken into consideration. There is little doubt that silica bodies have appeared in separate lines of evolution, as is indicated by their totally different shape in the Poales and the Cyperales. They are mostly lacking in the Juncales which are closely related to the Cyperales and in some other respects have a more basic organization. In the Cyperales they generally form conical bodies with the tips of the cones directed outwards, while in grasses they have various, but not conical, shapes. By considering the distribution of numerous such attributes, a more profound knowledge of connections between taxonomic groups of higher categories may be possible. PROSPECTIVES, RELATION TO A FEW OTHER CURRENT CLASSIFICATIONS There is a tendency at present for the current systems of classification of angiosperms to approach one another. Insufficiencies have become obvious and

30 120 R. M. T. DAHLGREN have been emended. Intensive studies result in clarification, as with the position ot Theligonum (Wunderlich, 1972). The systems of Takhtajan (19691, Cronquist (1968, 1979) and Thorne (1976) are at present under revision. The last version of Takhtajan s diagram may be that shown in Jones 8c Luchsinger (19791, but a fuller version is soon to be published in the Botanical Review (Cronquist, personal communication). Takhtajan (19691, like Cronquist (19681, uses the subclass level as an intermediate rank above the order and superorders. In the monocotyledons, the Arales and Typhales are associated with the Arecales, Pandanales and Cyclanthales to form the Arecidae. In the present classification the former two orders are not allied to the latter three nor are they allied with each other. With certain exceptions the main grouping of the monocotyledons is otherwise in good agreement with the present system. In the dicotyledons the subclass Magnoliidae largely corresponds to the Magnolianae and Nymphaeanae of the present classification. A closer association between the Piperales and Nymphaeales is here supported by the similarity of seed. In the Ranunculidae of Takhtajan (1967) the Sarraceniales no doubt form an alien element and would be better associated with Ericales or Cornales (Dahlgren, 1975a; Jensen et al., 1975). In Takhtajan s Caryophyllidae there are doubts about the affinity between the three orders Caryophyllales, Polygonales and Plumbaginales. The Urticales in Takhtajan s system are associated with the Hamamelidae, while most modern data seem to support a connection with the Malvales- Euphorbiales-Thymelaeales lines treated by Takhtajan in the Dilleniidae. It seems questionable that the Hamamelidales should appear in so remote a position from the Rosales- Saxifragales line of the Rosidae. The Dilleniidae in Takhtajan s classification include the order Ericales, which is thus shown as an evolutionary branch very distant from the Cornales. This is not supported by embryological and chemical data and is a weak point shared also by the systems of Cronquist and Thorne. Further points to be mentioned are the association of the Nepenthales and Elaeagnales with Saxifragales and the close association of the Araliales with the Cornales, which is again not supported by embryological and chemical characters. Furthermore, the Araliales are very remote from the Asterales (and Campanulales) in Takhtajan s system, which is not in accord with recent chemical and other data (CJ: Hegnauer, 1964). Otherwise, most groups of the iridoid plants (except Ericales and Cornales) are fairly closely placed, although interrupted by the iridoidless Polemoniales. In its main features the latest version of Cronquist s system (Cron uist, 1979; diagram in Jones 8c Luchsinger, 1979) is very reminiscent of that o 9- Takhtajan ( 1969) and recognizes the same subclasses, although the Ranunculidae are included in the Magnoliidae; an additional superorder was recently erected in the monocotyledons, namely the Zingiberidae (comprising the Bromeliales and Zingiberales). Among the Commelinidae the Cyperales include the grasses (Poaceae) as well as the Cyperaceae which contradicts morphological, cytological, anatomical and other evidence. The inclusion of Typhales in Commelinidae is in accord with most attributes of this order, although the wind pollination syndrome responsible for this may have evolved independently. The Magnoliidae correspond to the Magnoliiflorae, Nymphaeiflorae and Ranunculiflorae of the present classification. The Piperales and Nymphaeales are more remote from each other (see above) than might seem justified from seed

31 \Z(,IO5PF,KM CLASSIFICATIO V 121 construction. The Caryophyllidae have the same circumscription as in Takhtajan s system and basically so have the Hamamelidae, although the small groups Eucornmiales, Didymelales, Leitneriales and Daphniphyllales form separate orders within this superorder in the Cronquist classification. The Dilleniidae, with the core group Theales in Cronquist s system also include the Ericales (and Diapensiales) which are thus placed far from the Cornales. The Nepenthales are associated with the Theales in Cronquist s system, as in the present classification, but include the Sarraceniaceae, the latter no doubt an alien element (though close to the Ericales). The Rosidae in Cronquist s system show less connection with the Dilleniidae than in Takhtajan s system but contain essentially the same orders. The Euphorbiales are associated with Celastrales in Cronquist s Rosidae, which is probably less desirable than their association with Mahales- Thymelaeales d5 in Takhtajan s system. The Apiales (= Araliales) in Cronquist s Rosidae are connected with the Sapindales but not (as in Takhtajan s system) with Cornales, which are placed in the same subclass; the remote position of the Apiales and Asterales is thus a common feature of both classifications. The Gentianales (separate from Rubiales) have a basic position in Cronquist s Asteridae where they separate the Asterales from the Campanulales, each of which being connected with various iridoid-containing orders. From a chemotaxonornic point of view this seems unwarranted. The iridoidless Solanales are in the midst of a number of iridoid orders, which comprise the Lamiiflorae in the present classification. In addition to the division of the angiosperms into 11 subclasses, which is common to the Takhtajan and Cronquist systems, both diagrams give the impression that extant orders are evolved from each other, although the angiosperms of Cronquist s system are ultimately rooted in hypothetical proangiosperms. The classificatory diagram of Thorne (1979) published as a separate chart, differs only slightly from a previous version (Thorne, 19781, likewise published separately at first but later published in Jones & Luchsinger (1979: 362). The superorders in this diagram radiate from an empty centre representing extinct protoangiosperms. As in the present classification, the superorders are more numerous than Takhtajan s and Cronquist s subclasses. The monocotyledon consist of the Commeliniflorae, Ariflorae, Typhiflorae, Areciflorae, Alismatiflorae, Triuridiflorae and Liliiflorae, the first and last mentioned of these being separated from each other by the other orders. This is unfortunate considering the close connections between the Liliiflorae with starchy endosperm (Velloziaceae, Haemodoraceae) and the Philydraceae, Bromeliaceae etc. placed in Thorne s Commeliniiflorae. The Zingiberales are included in the Commeliniflorae, joining up with the suborders Bromeliinae- Ponterderiineae (-Commelinineae) of the Commelinales, which is in good agreement with the views held here although the Bromehneae and Pontederiineae form separate orders of my Liliiflorae. The Chenopodiiflorae in Thorne s system of 1979 are liberated from the Polygonales and Plumbaginales and thus are well defined, while the Theiflorae is the more extensive, including the Ericales, Primulales (with Plumbaginaceae), Polygonales and Ebenales. The Ericales and Sarraceniaceae (included in the Theales) thus are widely separated from the taxa of Cornales as circumscribed here. Thorne does not pay any attention to the correlation between the embryological attributes and iridoids as outlined above

32 122 R. M. T. DAHLGREN (they are not considered by Takhtajan or Cronquist either) and the families of the Corniflorae, Loasiflorae, Gentianiflorae and Lamiiflorae are therefore scattered in his diagram, which for this reason appears to be unnatural in some sections. However, the Oleales have been placed in conjunction with the Gentianales, and the last classification (Thorne, 1979) shows the Solanales-Campanulales (Solaniflorae) in close conjunction with the Asteriflorae, which is also approached by the Araliales, all this being in accord with recent findings. Chemotaxonomically and embryologically unnatural units are the Corniflorae in Thorne s system which consist of the Araliales, Dipsacales and Cornales, the last-mentioned of which include Vitaceae. According to some evidence, the Solanales of Thorne s superorder Solaniflorae agree better with the Boraginales (of Thorne s Lamiiflorae) than with Campanulales with which they are associated by Thorne. The Lamiales, which are associated with the Boraginales in Thorne s Lamiiflorae, would also match better with the Bignoniales of his Gentianiflorae, where Bignoniales are placed together with Gentianales-Oleales. Fabaceae Jemu luto is treated in the suborder Fabineae together with the Sapindineae and Rutineae in Thorne s Rutiflorae. This is a radical step which gains support from various sources. On a whole, the classifications of Thorne ( ) are characterized by fewer and more widely circumscribed orders than in the present system. However, the orders are frequently divided into several suborders. The families are likewise relatively widely circumscribed, but often divided into fairly distinct subfamilies. The hierarchy thus established is somewhat heavier than in other contemporary systems, but allows a more modulated picture of interrelationships. In nomenclatural construction the classification presented here agrees best with that of Thorne, as the subclass level has been reserved for the mono- and dicotyledons, and the superorder level has been widely exploited. The superorders of the classifications are also of comparable number and size to those in Thorne s classification and I have adopted his nomenclatural construction, using the termination --orae for the superorders. This ending has been extensively used for a long time, especially for some groups of orders, e.g., the Liliiflorae and Rosiflorae. It is to be hoped that some decisions of a purely nomenclatural character and of rank can be taken in order to give the current classifications (at least superficially) a more comparable appearance. Divergent philosophy, and the interpretation and different use of factual information will inevitably continue to give each classification individual features, and it would be wrong to violate sound conclusions merely in order to achieve an increased correspondence between the different classifications. REFERENCES BEHNKE, H. -D., Die Siebrohren- Plastiden der Monocotyledonen. Vergleichende Untersuchungen uber Feinbau und Verbreitung eines characteristischen Plastidentyps. Phnta, 84: BEHNKE, H. -D., P- und S-Typ Sievelement-Plastiden bei Rhamnales. Eeitriige zur Etologie der PJanren, 50: BEHNKE, H. -D. & DAHLGREN, R., The distribution of characters within an angiosperm system. 2. Sieve-element plastids. Eotaniska Notiser, 129: BERG, C. C., I97 7. Urticales, their differentiation and systematic position. Plant Systematics and Evolution, Supplement I: BOULTER, D., The molecular evolution of higher plant cytochrome c. In T. Swain (Ed.), Chemistty in Euolution and Systematics. London: Buttenvorths & Co.

33 .\S(;IOSPER~l CLASSIFICATIO Y I23 BOUMAN. F. Re CALIS, J. I. M., Integuinentan shifting-a third wav to unitepi\. Berichle ilrr Deutrchen BofnnisLhen GesuIlrrhuJ, , BREMER, K. & WANNTORP, H. -E., Phylogenetic svstematics in botany. Taxon BRIGGS, B. G. Re JOHNSON, L. A. S., Evolution in the Mvrtaceae-evidence ti-om inhoi-rscence wii( tuw. Prorefdings ofthe Linnean Sociel). of,yew South Wales, 102: CORNER, E. J. H The Serds ofdicotykdons. Cambridge: Cambridge University Prrss. CRONQLJlST. A The Evolution and Classification of Flowering Plants. London & Edinburgh: Nelson & soll5. CRONQIIIST, A Uulline of Classlficalion ofthe Living Families of Seed Plantr Houj in Knurli the Srrd Plantr: Dutiiiqur, Iowa' Win. C. BrownCo. CIJLVENOR, C. C. J Pvi-rolizidine alkaloids - occurrence and systematic importanre in angiorpermr. Hotaiii\kn,Vott,er. 131: DAHLGREN, R., I97ia. A \\stein of classification of the angiosperms to be used to driiionstratr thr disti-ibution olcharacters. Botanisha Notiser, 128: DAHLGREN, R.. 197ib. Thr distribution of characters within an angiosperm system. I. Soiiir embryological chi-artrrs. Hotanisha.Voliser, 128: DAHLGREN. R., (in cooperation with B. Hansen. K. Jakobsen & K. Larsenl. Angiospermerner Taxfinomi, 4 Kdmihavn: Akademisk Forlag. DAHLGREN, R. 1977a. A cominentary on a diagrammatic presentation ofthe angiosprrnmr in relation to the distributioii of c-haractei- states. Plant Syslemattcs and Evolution, Supplemen1 I : DAHLGREN, R., 1977b. A note on the Taxonomv of the 'Synpetalae' and related gi-oupb. Puhiitatioizi ofthe Cairo ljniiier\iti' Ilerbarium, i-x: DAHLGREN, R. Re CLIFFORD, H. T., in press. The monocotvledons. A comparativr 5tudv. Rolanzra/ S~~temaltrr, 2. London: Academic Press. DAHLGREN. R., NIELSEN, B. J., GOLDBLAIT, P. & ROURKE, J. P., Further note\ on Rrtriaccar, its c-hrinic,il (oiitriitr and aflinities. Annals $the Missouri Botanical Garden, 66: DAHLGREN, R., JENSEN, S. R., & NIELSEN, B. J., Iridoid compounds in Fouquieriarrae and note5 on it) possible dfhnities. Bolanishn,Voliscr, 129: ENDRESS, P., Gesichtpunkte zur systernatischen Stellung der Eupteleaceen (Magnolialesi. Brrichle drr Schturizeri~hen Hotuntrchen Gesellschaft, ENDRESS, P. K., Evolutionary trends in the Hamamelidales-Fagales Group. Plant.Tv\lemnficr and Evnlzitiun,.Supplement I: GORNALL. R. J., BOHM, B. A & DAHLGREN, R., The distribution of flavonoidc in the angiosperms. Hotani\ha.Votisei. 132: 1-30, HANSEN, B., The genus Aalanophora J. R. & G. Forster, a taxonomic monograph Dnnsk Botanish Arkiu, 28. HEGNALJER, R., Cheinotaxonomie der Pflanzen, 3. Basel & Stuttgart: Birkhauser Verlag. HEGNAUER, R., Chemotaxonomie der Pflanzen, 5. Basel & Stuttgart: Birkhauser Vrrlag. HEGNAUER, R., Chemolaxonomze der Pflanzen, 6. Basel & Stuttgart: Birkhauser Verlag. HUBER, H., Die Saineninerkmale und Venvandtschaftsverhaltnisse der Liliifloren..2.lillrifungen der Bolanirchrn Statr\ammlung Miinchen,U: HUBER, H., 1977 The treatment of the monocotyledons in an evolutionan rvsteni of ch\sification. Plant Syctematicr and t'i*olution. Supplemrnt I : JAGER-ZURN, I., Zur Frage der systematischen Stellung der Hydrostachydceae auf G~-und ihrer Embryologie, Blhten- und InHorescenz-inorphologie. i)sterrezchiiche Botanische Zeitschrifl, JENSEN, S. R., NIELSEN, B. J., & DAHLGREN, R., Iridoid compounds, their occurrence and systematic importance in the angiosperms. Botaniska Nottser, 128: JONES, S. B. & LUCHSINGER, A. E., Plant Syslemalrcs. New York: McGraw Hill. MABRY, T. J. & BEHNKE, H. -D., ieds) Evolution of Centrosperrnous Families. Plant SJ\lmaticr and E~iiolution, 126: PHILIPSON, W. R Ovular morpholog\ and the major classification of the di( orvledons. Eotantcal Journal ofthe Linrienn Societv RAO, 7. A. & DAS. S Leafsclereids - occurrence and distribution in the angiosperms. Botanl\ka.Votzsrr. 132: ROMEIKE. A,, Tropanr alkaloids - orcurrence and systematic importance in angiospernia. Aolnntsha A'oticer, 131: STEBBINS, G. L., Flowering Plants. Evolution above the Spectes Level. Cambridge, Massachusetts: Belknap Press of Harvaid l~niversirv Press. TAKHTAJAN, A,, Flowering Planls. Origzn and Dispersal. Edinburgh: Oliver & Boyd. THORNE, R. F., Svnopsis of a putative phylogenetic classification of the floweringplants. A h, 6' THORNE, R. F., The 'Amentiferae' or Hamamelidae as an artificial group: a summan Ytatement. Brittonta, 25 ' THORNE, R. F., IS73 Inclucion of the Apiaceae (Umbelliferae) in the Araliaceae. "v'oter[rnm the Hoyd Botanic Garden Edinburgh, 32: THORNE, R. F., A phylogrnetic classification of the Angiospermae. Evolutiona? Biolog),