Botanical Journal of ike Linnean Society (1991), 107: 1-34. With 54 figures

Leaf anatomy of : ecological, systematic and phylogenetic aspects

SHERWIN CARLQUIST, F.L.S. • Rancho Santa Ana Botanic Garden and Department of Biology, Pomona College, Claremont, California 91711, U.S.A.

Received August 1989, accepted for publication October 1989

CARLQUIST, S., 1991. Leaf anatomy of Bruniaceae: ecological systematic and phylogenetic aspects. Quantitative and qualitative data are given for 60 species of the 12 genera of Bruniaceae; most data are based on liquid-preserved material. Leaves of Bruniaceae are basically linear (broader forms are probably derived) with an apicula that contains phellogen activity. Most bruniaceous leaves have some degree of isolateral construction, with transition to normal bifacial construction in a few species, but more commonly transition to 'inverse' bifacial structure (stomata on adaxial face, palisade on abaxial face). The latter type is correlated with the tendency for leaves to be appressed to stems. Tannins and very likely other dark-staining materials are very characteristic of mesophyll cells. Six genera have a large strand of fibres on the midvein and rhomboidal crystals in bundle sheath cells. The other six genera have few or no fibres on veins and have druses in mesophyll cells (but not in bundle sheath cells;. These distinctions may relate to intrafamilial , but they also support the primitive position usually accorded to , Thamnea and Tittmannia. A key to genera based on leaf antomy is offered. Details of epidermal cell shape, cuticular relief and trichome form and structure based on scanning electron microscopy are given. Leaf anatomy, combined with other features, favours a relationship between Bruniaceae and Grubbiaceae in particular and in broader contexts allies Bruniaceae to rosalean and possibly hamamelidalean families.

ADDITIONAL KEY WORDS: —Ecological anatomy - Grubbiaceae - Hamamelidales - Pittosporales - Resales.

CONTENTS Introduction 2 Material and methods 3 Anatomical descriptions 3 Leaf tip 3 Leaf shape and dorsiventrality 7 Cuticle 11 Waxes 16 Epidermal cell shape 18 Subsidiary cells 18 Guard cells 18 Trichomes 20 Contents of mesophyll and epidermal cells 24 Crystals 25 Veins 25 Ecological conclusions 27 Systematic conclusions 29 Phylogenetic conclusions 32 Acknowledgements 33 References 33 1 0024-4074/91/090001-1-34 S03.00/0 •© 1991 The Linnean Society of London 1 S. CARLQUIST

[NTRODUCTION Bruniaceae comprises 12 genera and about 75 species (Bond & Goldblatt, 1984). These small to large are confined to south-westernmost Africa; most species are characteristic of the montane sandstone areas, but a few can be found on lowland areas and even on laterites. Only one species, Raspalia trigyna, exceeds the borders of Cape Province (Pillans, 1947). The leaves of Bruniaceae are predominantly small and either acicular or scale­ like, although broader leaves occur in species of Lonchostoma as well as in cordifolia and Pseudobaeckea cordata. The limited surface area of leaves in most species of Bruniaceae seems related to particular features of the Mediterranean-type climate of Cape Province. The xeromorphy reflected in these leaves is indicative not so much of low annual rainfall but of high summer heat, and low summer humidity. Montane Cape Province is notably windy and sandstone soils are very porous. Nevertheless, during field work in South Africa, one repeatedly sees that Bruniaceae tend to occur in mesic microhabitats: south- facing slopes, seeps, marshes, streambanks, and in the shade of boulders. These microhabitats, together with the leaf xeromorphy of the family, may explain how the relatively primitive woods that Bruniaceae have can persist in a Mediterranean-type climate (Carlquist, 1978a). Small leaf surfaces minimize transpiration, but are adaptive because of the sunny climate and lack of shade- offering trees in localities where Bruniaceae grow. However, the leaves of Bruniaceae are not uniform, and this diversity can be related to the diversity of habitats in which species of the family are found. Features of leaf structure have been explored by various authors, notably Colozza (1904), Kirchner (1904), Marloth (1925)' and Niedenzu & Harms (1930). The present account amplifies knowledge of leaves of the family because it is based on liquid-preserved specimens of the majority of species, collected during a visit in 1973. Leaves of dried specimens have supplemented the liquid- preserved specimens. Data from the above-mentioned collections are sufficiently comprehensive to enable distinctions to be made to and species concepts within the family. The monograph of Pillans (1947) forms a convenient framework for that comparison. In addition, recent data from wood anatomy (Carlquist, 1978a), cytology (Goldblatt, 1981) and palynology (Hall, 1988) can be incorporated. Leaf anatomy proves to offer more characteristics than one might think interpretable at the level of subfamily or tribe, and the subfamilial groupings offered by Niedenzu & Harms (1930) and Takhtajan (1987) are pertinent in this regard. A few data from leaf anatomy appear to be interpretable with regard to the phylogenetic sequence of genera within the family. Although data from leaf anatomy might be expected to be applicable mostly to infrafamilial taxonomy, a few foliar features seem interpretable at the interfamilial level. Two main lines of relationship have been claimed for Bruniaceae: ericalean (Dahlgren & Van Wyk, 1988) and rosalean (Van Tieghem, 1897; Thome, 1976; Dahlgren, 1980; Cronquist, 1988). Grubbiaceae has been claimed to have various affinities, including ericalean (Fagerlind, 1947; Cronquist, 1988), and this complicates the affinity found by some authors between Bruniaceae and Grubbiaceae (Carlquist, 1977). LEAF ANATOMY OF BRUNIACEAE •;

MATERIAL AND METHODS Liquid-preserved leaves were embedded in paraffin and sectioned according to the usual techniques. Dried leaves were treated overnight with warm 2.5% aqueous NaOH, stored in aqueous 50% ethyl alcohol, and then embedded and sectioned as with liquid-preserved specimens. Both transections and paradermal sections were cut for all species. Sections were stained with a safranin fast green combination. Some mesophyll cells stain very deeply because of dense contents of tannins or another, possibly terpenoid, compound, or both. These sections are opaque to a large extent, and do not photograph well. Nevertheless, sections have been illustrated photographically here in order to show accurately the nature of these accumulations. Leaves of most species were examined by means of scanning electron microscopy (SEM). Dried leaves were coated with gold and examined with an ISI WB-6 SEM. Coverage of the family is extensive, considering that 21 species are listed as critically rare, threatened or extinct (Hall & Veldhuis, 1985). Material of some species was not available. Berzelia (12 species) is less well-covered because preliminary investigations showed it to be more uniform; species in that genus and in were selected for study on the basis of texture and shape diversity. The monograph of Pillans (1947) has been followed with certain exceptions. Author citations are listed in Table 1. Lonchosloma esterhuyseniae was added by Strid (1968) and Tittmannia esterhuyseniae was named by Powrie (1969). On the advice of E. Powrie and of E. Esterhuysen (personal communication), I have differed from the treatment of Pillans (1947) in certain nomenclatural matters. Linconia deusla Pillans is considered a synonym of L. cuspidata. In Slaavia, S. pinifolia is considered the correct name for S. dregeana Presl; S. capilella is considered correct for S. comosa Colozza, and S. trichotoma Pillans is a synonym of S. capilella. Berzelia squarrosa is considered a valid segregate of B. arachnoidea. My collections are represented in the herbarium of the Rancho Santa Ana Botanic Garden (replicates distributed to numerous other herbaria). Collections of other workers are located in the Compton Herbarium of Kirstenbosch Botanic Garden. Length measurements in Table I are taken from Pillans (1947) or from actual specimens and are given to the nearest 0.5 mm. Width and thickness measurements are given to the nearest 0.01 mm, and are taken from microscope slides. All quantitative data in Table 1 represent conditions judged to be typical, but in no instance was sampling that could lead to statistical significance undertaken. Nevertheless, contrasts sufficiently extreme to possess significance are claimed in the data presented. Genera and species are arranged alphabetically in both Table 1 and in the plates.

ANATOMICAL DESCRIPTIONS Leaf tip The entire family Bruniaceae is characterized by distinctive brownish apices termed apiculae here. Metcalfe & Chalk (1950) and Dahlgren & Van Wyk TABLE 1. Leaf characteristics of Bruniaccac

Character k

Species Collection 1 2 3 1 5 6 7 8 9 10 U 12 13 14 15 16

Auduuima capitata Brongniart Carlquist 4460 5.5 0.80 0.41 4 'i 1 5 4- + St K 7 a 2.2 0.17 Berzelia abrolanoides Brongniart CarUpdsi 4478 3.0 1.14 0.52 10 10 2 3 + + s !< 1 in 1.7 0.09 B. arachnuidea F.cklon & Zeyher Carlquist 4646 12.0 0.90 0.70 9 5 2 2 5 + + s K 5 in 5.4 0.68 B. comma lata Sunder Carlquist 4736 5.0 0.53 0.36 5 7 1 6 + + S R 4 a 1.3 0.11 B. cordifolia Schlectendal Carlquist 4690 7.1) 4.50 0.73 6 0 1 6 + 4- S R 18 A 15.8 1.32 B. ecklonis Pillans CarUpdsi 4963 4.5 0.92 0.51 3 6 1 4 +- i- s R 5 .i 2.1 0.23 B. matron Pillans Estnhuysen 4984 6.5 0.89 0.43 11 14 1 4 + + s R 5 a 2.9 0.12 B. rubra Schlechtendal Carlquist 4967 11.0 1.40 0.63 16 18 1 9 + + s DR 7 a 7.7 0.23 B. squarrosa Thunberg Carlquist 4945 7.0 0.87 0.56 11 11 1 4 4- + s R 6 a 3.0 0.14 Bruma albiflora Phillips Carlquist 4576 11.0 0.60 0.33 7 7 1 4 + + s R 3 a 3.3 0.24 n B. alopecuroides Thunberg Carlquist 4672 3.5 0.97 0.57 5 0 1 1 0 4- s R 4 a 1.7 0.15 > B. macrocephala Willdenow Si,,!.,,,- 61357 8.5 0.86 0.55 7 8 2 4 0 4- S R 7 a 3.7 0.24 H B. neglntu Schlechter Eslerhuysen 4558 4.0 0.68 0.50 7 11 1 4 4- 4- S R 2 a 1.4 0.08 Linconia alopecurnidea L. Taylor'4247 18.0 1.14 0.57 6 6 1 5 0 + s D 7 0 10.3 0.86 /.. aispidala Sw.iu/ Carlquist I'.ll J 4.0 1.23 0.55 6 6 1 5 II 4- si D 5 0 2.5 0.21 Lonchostoma esterhuyseniae Slrid Carlquist 4982 5.0 2.10 0.31 11 18 0 2 3 + 0 s l) 5 M 5.3 0.18 /.. monogyna Pillans Carlquist 4682 5.0 4.20 0.27 2 7 0 2 3 4- 0 s D 14 f 5.3 0.58 /,. myrtoides Pillans Carlquist 5020 11.0 7.30 0.43 5 9 0 2 3 + 0 s 1) 21 M 40.2 2.87 L. pentandrum Pillans Carlquist 4891 8.0 4.10 0.28 1 9 0 2 3 + 0 s 1) 17 F 16.4 1.26 L. purpureum Pillans Carlquist 4930 5.5 3.50 0.34 6 11 o 2 3 + 0 s li 13 \1 9.6 0.56 Mniothamnea bullata Schlechtei Carlquist 4715 2.0 1.38 0.31 2 5 o 1 4 + 0 Si 0 :> f 11 0.20 .11. lallunoides Niedenzu Carlqtdsi 4720 1.5 0.77 0.27 1 6 0 1 3 4 ii s D 5 0 0.6 0.09 Mebelia fragarioides 0. Kze Carlquist 4934 4.0 0.57 0.47 3 7 2 3 4- 0 SI DR 2 a 1.1 0.11 jV. laevis (). K/.< Rourke 709 8.0 0.95 0.52 20 20 1 9 4- 0 SI R 5 a 3.8 0.10 .V. paleacta Sweet Carlquist 4780 3.5 0.75 0.44 3 11 1 4 + 4- ST R 4 A 1.3 0.09 A sphaerocephala O. Kze Cmlquist 4780 4.5 0.68 0.36 3 7 1 4 + ii ST R 3 A 1.5 0.15 .V. stokuei Pillans Carlquist 5022 4.5 0.98 0.60 5 7 '_' 2 4 -I + ST R 4 A 2.2 0.18 Pseudabaeckea qfricana (Burin, I'd.) Pill. Carlquist 4782 1.5 1.23 0.41 5 7 1 3 + + S 1) 8 r 0.9 0.08 P. cordata Niedenzu Carlqtdsi 4716 8.0 1.26 0.22 3 1 1 2 f 4- ST D 8 0 5.0 1.25 P. cordata Carlquist 4928 7.0 1.67 0.29 3 4 1 2 + -t si D 10 0 5.8 0.83 P. slokoei Pillans Stokoe 16886 6.0 2.46 0.22 7 1 0 3 0 4- SI 1) 11 0 7.4 0.93 P. teres Dummer EsUrkxrsa 33463 1.5 0.65 0.27 2 16 0 2 4 4- 0 s R 3 a 0.5 0.03 Raspalia bamardii Pillans Barnard X-1925 6.5 1.14 0.36 6 7 0 3 + (1 Si 1) 8 1 3.7 0.2)1 R. drtgeana Niedenzu Esterhuysen 7589 3.0 0.85 0.31 5 9 (I 3 + II s n 9 0 1.3 0.09 R. globosa (Lam.) Pill. Cailquist 4811 5.0 0.93 0.40 5 14 0 4 i II s 1) 5 2.3 0.12 R. microphylla Brongnian Carlquist 4673 1.8 0.62 0.27 5 27 0 3 + 0 s 1) 1 0 0.6 0.02 R. ohlimgifolia Pillans Esterhuysen 29883 4.0 0.94 0.32 7 11 1 3 + II Si 0 7 1.9 0.10 R. palustris (St hl<. i< r Pill. StokOi 56829 4.0 0.84 0.22 2 3 1 2 + 0 S 1) 3 0 1.7 0.34 R. phylieoides Aniott Carlquist 4940 2.0 1,46 0.36 3 11 0 3 + II s 1) i) 1.5 0.11 A', uucukta (Bolus) Pill. Esterhuysen 12458 4.5 0.73 0.19 3 0 3 0 p 1) 6 0 1.6 0.32 R. slaavioides (Sonder) Pill. 2 + Stokoe 5681 6.0 0.83 0.43 5 7 1 5 + + Si 1) 4 2.2 0.18 R. stokoei Pillans Carlquist 5053 3.0 1.46 0.47 7 14 1 4 0 SI 1) 7 0.9 0.04 R. trigyna Diimmcr • Taylor 2636 1.0 0.42 0.23 3 12 0 3 + II S II 3 II 0.2 0.01 R. variabilis Pillans Carlquist 4936 1.5 1.14 0.44 3 in 1 5 II St 1) 3 1.3 0.10 R. nllosa Presl + I'owrie 137 1.5 0.63 0.23 2 1 2 + 0 Si 1) 1 n.-. 0.10 R. virgata (Brongniart) Pill. :( Carlquist 4936 2.5 1.14 0.29 1 ii 1 ) 3 0 0 1.4 0.12 brownii Diininici + s / Stokoe 66971 1.03 0.5.'! 14 1 7 0.18 s Kipitella Sonder B.0 9 5 + + s, 1) 4.1 Carlquist 4819 1.18 3.5 0.17 S. dodii Bolus 6.0 0.51 9 12 1 I 4 + + s 1) 7 S. glutinusa Dahl Carlquist 4620 8.0 2.02 0.49 in 21 1 5 + + s D 12 8.1 0.20 S. pinifolia Willdcnow Carlquist 4768 12.0 0.75 0.54 7 9 1 5 + + s D 5 4.5 0.46 .V. radiata Dahl Carlquist 4754 6.5 1.15 0.79 7 II 1 8 + 4- s 1) 8 3.7 0.32 S. verticillala (L. fil.) Pill. Carlquist 4601 5.5 0.75 0.37 9 18 1 4 4- + s 1) 5 2.1 0.08 .S'. zeyheri Sonder I'owrie 58 3.0 0.68 0.55 2 5 1 5 + + ST D 5 1.0 0.13 lhamnea diosmoides Olivci Wiliams 752 15.0 1.10 0.48 25 If> 1 5 + + ST D 6 f 8.3 0.20 T. hirtella Oliver Carlquist 4665 1.5 0.28 0.22 5 12 1 4 0 + St DR 3 A 0.2 0.01 T. massoniana Diimmcr Esterhuysen 25765 2.5 0.58 0.36 Id 11 1 4 0 + St R 3 a 0.7 0.03 7. thesioides Diiminer Carlquist 4852 3.5 0.78 0.46 111 12 1 4 0 4- St R 5 A 1.4 o.ot. / ittmannia esterhuyseniae Powric Mm Owen 37624 1.0 0.49 0.31 8 9 1 4 0 4- •) R 3 a 0.2 0.01 /. lu.spula Pillans Carlquist 5040 5.0 0.75 0.47 7 25 1 5 + 4- St R 7 a 1.9 0.06 T, laevis Pillans Stokoe 56818 4.0 0.39 0.28 5 5 1 0 5 0 4- S R 4 a 0.8 0.08 T. laxa Presl Carlquist 4850 2.5 0.39 0.32 3 Ii 1 4 + 4- s R 3 a 0.5 0.05 Carlquist 5051 4.0 0.33 0.33 5 •i 1 4 + 4- ST R 3 a 0.7 0.05

•Key 1, Leaf length, mm; leaf width, mm; 3, leaf thickness, mm; 4, adaxial cuticle thickness, urn; 5, abaxial cuticle thickness, urn; 6, number of udaxial palisade layers; 7, number of abaxial palisade layers; 8, number of spongy layers; 9, stomatal presence Oil adaxial surface ( + = present, 0~ absent); 10, sioniaial presence on abaxial suil.n e (4- "present, 0=absent); 11, mesophyll cell contents (S = reddish-staining deposits i" palisade plus at least some spongy cells; s = reddish-staining deposits in some palisade cells, leu spongy cells; T••tannins in essentially all mesophyll cells; t = tannins in some mesophyll cells in which reddish-Staining deposits are not also present); 12, crystal presence (D = druses in mesophyll cells; R = rhomboi(lal crystals in bundle sheath cells; 0 = no crystals observed); 13, number of veins and veinlets as seen i" a transection; 14. fibre presence on veins la = large strand of fibres; f= Strand of few fibres; m~ moderate number of fibres; upper iase = strands present on several veins, lower case = fibre strands only on midvein; 0 = no fibres observed); 15, leaf area (length x width divided by 2), mm'; 16, mesomorphy index leaf area divided by sum of adaxial and abaxial cuticle thickin S. CARLQUIST

Figures 1-4. Leaf anatomy of Audouima and Berzelia. Figs 1-3. Audouima capilala {Carlquist 4460). Fig. 1. Transection, adaxial face at left, showing the large strand of fibres adjacent to phloem <>l the midvein. Fig. 2. Portion of mid veins Fibres are gelatinous; rhomboidal crystals are in bundle sheath cells, above. Fig. 3. SEM photograph of leaf margin; distally pointing trichoma are shown (one at left has lost tip portion). Fig. 4. Berztlia commulata [Carlquist 4730), margin of leaf showing stonuiia. epidermal relief, and papillate cells (right). Scale bars: Fig. 1 = 21 pm, Fig. 2 = 76 pm, Fig. 3 = 20 pm and Fig. 4 = 88 pm.

; 1988) state that this apicula is absent in Audouinia, but this is not true: the apicula is present on young leaves of Audouinia, but falls off relatively early. This leaf tip is more than a mucro. Some of its peculiarities are detailed by Niedenzu LEAF ANATOMY OF BRUNIACEAE 7 & Harms (1930: 293). The tip is cellular, as can be seen here in SEM pictures (Figs 10, 16, 20, 25, 34, 45, 48) and in sections (Fig. 51), but the cell walls are very thin. Niedenzu & Harms state that the apex of the apicula is essentially just the tip of the leaf, but the base of the apicula is derived from phellogen activity ("spater verkorkt"), and this agrees with my observations. The majority of tissue in the apicula is not produced by the phellogen. The rows of cells that show the action of the phellogen (Fig. 51) are present in a third or less of the apicula. The apicula is appreciably shrunken compared to subadjacent leaf portions (Figs 10, 16, 20, 34, 48), rarely only slightly shrunken (Fig. 45). Its surface may be smooth (Figs 10, 48), but is more commonly rough (Figs 25, 45). When smooth, cells of the apicula are covered by an intact cell wall and cuticle. Roughened apicular surface indicates a natural fracturing of the epidermal cell wall including its cuticle, and may reveal (Fig. 45) globular cell contents that are considered possibly massive deposits of terpenoids. These deposits give a brownish colour to the apicula. Although such deposits are abundant elsewhere in leaves of Bruniaceae, they probably do not confer a brown colour to other leaf portions because the deposits are located in subepidermal layers; presence of thick cuticle may also mask this colouring. The cell wall and cuticle of the apicula are so thin that the brown colour would not be masked even if they are intact. In Staavia dodii, a thick layer of cuticle lying between the phellogen base of the apicula and the underlying storage tracheids of the leaf tip was observed. Such a cuticle, as well as the apparent occlusion of the leaf by the phellogen in all species, suggests that the apicula does not function as a hydathode.

Leaf shape and dorsiventrality Leaf shape in Bruniaceae is worthy of discussion here because it relates to whether leaves are isolateral or bifacial in construction. In general, leaves of Bruniaceae are described as linear. A comparison with conifer leaves is apt because in many cases leaves are longer (Table 1, column 1) than wide (Table 1, column 2), and are relatively slender, although ordinarily at least half as thick (Table 1, column 3) as wide. Leaves of Bruniaceae cannot be termed 'cricoid' if by that term one connotes recurved margins forming one or two pockets on the lower side of the leaf: leaf margins are not recurved in Bruniaceae (Figs 1, 8, 11, 21, 22, 35, 41, 48). In most species leaves are 1.5 to 2.0 times as wide as thick (compare columns 2 and 3 in Table 1). Notably wide leaves occur in Berzelia cordifolia, Lonchostoma esterhuyseniae, L. purpureum, Pseudobaeckea stokoei and some collections of P. cordala. Flat leaves with similar length-to-width proportions but with much smaller size characterize other species of Lonchostoma and the species of Mniothamnea, Raspalia and Thamnea; these leaves may be characterized as scale-like. Because the midvein in leaves of most Bruniaceae is much larger than the lateral veins, the shape of leaves as seen in transection is often triangular or triquetrous. This can be seen in mild form in Audouinia (Fig. 1), but is more pronounced in the species of Thamnea (Fig. 48) and most species of Raspalia. Leaves of Linconia alopecuroidea are highly distinctive in this respect. The abaxial surface (or two surfaces) are divided by a ridge that lacks stomata (Fig. 17, right); the margins are similar, but more pilose (Fig. 17, left). Between the margins and the abaxial keel, the two longitudinal surfaces bear stomata 8 S. GARLQUIST

Figures 5-8. Leaf anatomy of Berzelia. Fig. 5. B. commutata (Carlquist 4730). margin of leaf showing stomata (left! and blunted short uniscriate trichomes (right). Figs 6-7. B. cordi/olia iCarlquist 4690). Fig. 6. Portion of leaf transection, adaxial face above; stomata on both surfaces, fibres below larger vein. Fig. 7. SEM photograph of stomata and wax deposits on abaxial epidermis. Fig. 8. B. squarrosa [Carlquist 4945). transection of leaf margin; fibres are absent on smaller veins. Scale bars: Fig. 5 = 27 urn. Fig. 6 = 76 urn, Fig. 7 = 68 urn and Fig. 8 = 21 urn. I.KAF ANATOMY OF BRUNIACEAE 9

Figures 9 12. Leaf anatomy of finwin. Figs 9-10. B. albiflora (Carlquist 457b). SEM photographs. Fig. 9. Leaf tip, showing apitulum (trichoma are on nearby stem). Fig. 10. Bulbous cells from margin near leaf tip; wax deposits in depressions between cells. Figs 11-12. B. alopecuroides (Carlquist 4672). Fig. 11. Transection of half of leaf, adaxial face above. Fig. 12. Portion from leaf transection showing fibres near phloem of mid vein and rhoml>oidal crystal in bundle sheath, below. Scale bars: Fig. 9=15 p-m, Fig. 10 = 77 urn, Fig. 11 = 21 p.m and Fig. 12 = 76 urn. Ill S. CARLQUIST

Figures 13-16. SEM photographs of Brunia leaf surfaces. Fig. 13. B. laevis [Carlquist 4695), trichomes on abaxial leaf surface. Figs 14-15. B. macrocephala (Esterhuysen 35424). Fig. 14. Abaxial leaf surface, showing coiled trichomes and stomala. Fig. 15. Adaxial leaf surface; long trichome plus numerous smaller coiled trichomes. Fig. 16. B. neglecla {Esterhuysen 4959), tip of leaf showing rounded shape of cells. Scale bars: Fig. 13 = 30 pm, Fig. 14 = 24 pm. Fig. 15 = 25 pm and Fig. 16=10 pm.

(Fig. 17, centre). Such a ridge separating stoma-bearing bands is absent in L. cispidata (Fig. 20). Leaves somewhat quadrangular in sectional view characterize the species of Brunia (Figs 9, 11, 16) and Nebelia (Fig. 35). Terete LEAF ANATOMY OF BRUNIACEAE ll leaves, which have a circular outline in transection, were observed in Titlmannia laevis, T. laxa (Fig. 52), and T. hispida. Although leaves of most species of Berzelia appear terete in gross aspect, they are at least somewhat flattened as seen in sectional view. Three criteria can be used for designating a leaf as isolateral or bifacial: cuticle thickness (Table 1, columns 4 and 5); presence of palisade chlorenchyma on both surfaces (Table 1, columns 6 and 7; spongy tissue occurs in all species, column 8); and presence of stomata on both surfaces (Table 1, columns 9 and 10). Only in Thamnea is there a lack of positive correlation among the above- mentioned features: in Thamnea, stomata occur only on the abaxial surface yet the cuticle on that surface is slightly thicker than that of the adaxial surface. Abaxial cuticle thickness is nearly identical to adaxial cuticle thickness in some species of Berzelia, Raspalia and Titlmannia. Otherwise, abaxial cuticle thickness in any given species is greater than adaxial cuticle thickness (an interesting exception occurs in one of the two collections of Pseudobaeckea cordata. The greater thickness of the abaxial cuticle may be taken as an indicator that (with the exception of Thamnea as noted above) the leaves of Bruniaceae are mostly inverse to the 'typical' dicotyledon leaf in greater development of cuticle thickness and palisade on the adaxial surface. However, few of the Bruniaceae are truly bifacial on the basis of mesophyll histology or stomatal distribution; most are transitional between the 'inverse' bifacial condition and isolateral structure. The closest approach to normal bifacial leaves is found in Pseudobaeckea cordaia (Carlquist 4716), in which cuticle is thicker on the adaxial surface, the abaxial palisade layer is somewhat like spongy parenchyma, and stomata are somewhat more abundant on the abaxial surface (although there are some stomata on the adaxial surface). Palisade cells are longer on the abaxial sides of leaves than on the adaxial surfaces in most species of Bruniaceae. This tendency is shown in Figs 1, 11, 21, 22, 29, 32, 41. Audouinia (Fig. 1), Brunia (Fig. 11) and Staavia show a close approach to isolateral construction, whereas Linconia cuspidala (Fig. 21), all species of Lonchostoma (Figs 22, 29), both species of Mniothamnea (Fig. 32) and all species of Raspalia (Fig. 32) show the inverse bifacial condition. In these, the abaxial surface has a thick cuticle and palisade chlorenchyma but lacks stomata (or has very few). Very close approaches to isolateral leaf construction or else true isolateral leaf characterize the species of Berzelia (Figs 6, 8), Linconia alopecuroidea (Fig. 17; no stomata on the adaxial surface, however), JVebelia (Fig. 35; stomata on adaxial surface in JV. fragarioides) and Titlmannia (Fig. 52).

Cuticle The cuticle in leaves of Bruniaceae is clearly differentiated from the epidermal wall. The epidermal walls in leaves of Bruniaceae stain brightly with fast green, whereas the cuticle stains with safranin. The pale staining of the cuticle renders it poorly visible in some photographs (Figs 1, 6, 8, 11, 21, 22, 29, 52). It is relatively easily seen in Figs 32, 35, 36, 53. Cuticular relief is absent or minimal in Audouinia capitala, Berzelia commulala (Fig. 5) or Mniothamnea callunoides (Fig. 34). In the latter, the cuticle obscures cell 12 S. CARLQUIST

Figures 17-21. Leaf anatomy of Linconia. Figs 17-18. l.incunia alopecuroidea .Taylor 4247). Fig. 17. Abaxial surface of leal, irichome-bearing leaf margin at left, midrib at right. Fig. 18. Epidermis from midrib showing cuticular relief. Figs 19-21. Linconia cuspidata (Carli/uist 4943). Fig. 19. F.pidermal cells near tip snowing papillate shape, light cuticular relief. Fig. 20. Abaxial leaf surface, showing stomata. Fig. 21. Transection of leaf, adaxial face at right. Scale bars: Fig. 17 = 8 um, Fig. 18 = 81 um, Fig. 19 = 50 urn, Fig. 20 - 10 urn and Fig. 21 = 21 urn. LEAF ANATOMY OF BRUNIACEAE 13

Figures 22-25. Leaf anatomy of Lonchosloma. Figs 22-23. L. esterkuyseniae [Carlquist 4892,. Fig. 22. Transection of leaf margin, adaxial face at left. Fig. 23. Portion of leaf transection; dark-staining material in epidermal and subepidermal cells; druse to right of veinlet, which lacks phloem. Figs 24—25. L. myrtoides [Carlquist 5020). Fig. 24. Abaxial epidermis; waxy deposiis on outlines of the papillate cells. Fig. 25. Leaf tip, cells of apicula show breakdown. Scale bars: Fig. 22 = 21 urn. Fig. 23 = 76 urn, Fig. 24 = 68 urn and Fig. 25 = 18 urn. II S. CARLQUIST

Figures 26-30. Leaf anatomy of Lonchosloma. Figs 26-27. L. monogyna (larlquist 4682), SEM photographs. Fig. 26. Abaxial epidermis. Fig. 27. Adaxial epidermis. Fig. 28. L. pentandrum iCarlquisI 5061), adaxial leaf epidermis from paradermal section. Figs 29—30. L. purpureum (Carlquist 4936~), leaf transections, adaxial surface at left. Fig. 29. Portion showing midvcin. with a lew libres adjacent to vein. Fig. 30. Midvein of leaf; druse at upper left. Scale bars: Fig. 26 = 77 pm, Fig. 27 = 69 pm, Fig. 28 = 21 pm. Fig. 29 = 21 pm and Fig. 30 = 76 pm. LEAF ANATOMY OF BRUNIACEAE 15

Figures 31-34. Leaf anatomy of Mniothamnea. Figs 31-33. M. bullata iCarlquist 4715). Fig. 31. SEM photograph of outer leaf surfaces. Fig. 32. Leaf transection, adaxial face at left, showing midvein. Fig. 33. SEM photograph of cuticle on outer surface. Fig. 34. M. callunoides {Carlquisl 4720), SEM photograph of outer surface of leaftip. Scale bars: Fig. 31 = 34 urn. Fig. 32 = 21 urn, Fig. 33 = 149 urn and Fig. 34 = 29 pm. 16 S. CARLQUIST outlines and forms an even coating over the leaf surface. Cuticular relief is very slight in Linconia cuspidata (Fig. 19), Raspalia stokoei (Fig. 43), R. virgata (Fig. 46) and Staavia radiata (Fig. 47). Cuticular relief as inconspicuous as this cannot be seen with the light microscope. The cuticular relief of Berzelia commutata (Fig. 4) is close to the limits of resolution of the light microscope. Noteworthy distribution of cuticular relief was observed in Linconia alopecuroidea, in which it forms prominent longitudinal striation on the bulbous cell walls of the midrib (Fig. 18). Cuticular relief occurs in Linconia cuspidata leaves only on the papillate cells near the apex of the leaf (Fig. 19); otherwise, the leaf surface is smooth (Fig. 20). The margins of leaves tend to have more cuticular relief than the surfaces in a few species such as Berzelia commutata (cf. Figs 4 and 5) and Raspalia variabilis. Cuticular relief takes the form of longitudinal striations plus minute roughness on the bulbous epidermal cells of Brunia albiflora (Fig. 10). The cuticular relief of Mniothamnea bullata (Fig. 33) appears as lamellate rather than as a series of ridges. The most common type of cuticular relief in the family is a series of longitudinal ridges, more common on the abaxial leaf surfaces than on the adaxial surfaces (cf. Figs 26 and 27). This was observed in Berzelia squarrosa, Brunia albiflora (Fig. 10), B. laevis (Fig. 13), Lonchostoma esterhuseniae, L. monogyna (Figs 26, 27), L. pentandrum, Nebelia fragarioides, N. paleaceae, N. stokoei, JV'. sphaerocephala (Fig. 38), Pseudobaeckea cor data Carlquist 4928 (Fig. 40), Raspalia barnardii, R. globosa, R. stokoei (Fig. 43) and R. variabilis. In leaves isolateral or nearly so, cuticular relief may be about as pronounced on abaxial surfaces as on adaxial surfaces: this was observed in Tittmannia laevis. Cuticular relief more prominent on the adaxial surface was observed in Linchostoma purpureum, Pseudobaeckea cordata {Carlquist 4716), Raspalia barnardii, R. dregeana and Staavia capitella (only fine striations present in this last species). Note should be taken of the difference between the two collections of Pseudobaeckea cordata with respect to cuticular relief. Comparison of these two collections leads to the generalization that cuticular relief is more pronounced on the leaf surface more exposed to sunlight. There are a few exceptions, as in the two species of Raspalia cited earlier in this paragraph.

Waxes Waxes occur commonly on leaf surfaces of Bruniaceae. Where most conspicuous, the waxes occur as rodlike or flangelike strands, best shown here in Figs 24, 26, 27, 33 (a few), 40, 43, 46 and 49. Sometimes the waxes are more flakelike than rodlike, as in Figs 38 and 40. Where epidermal cells are dome-like in shape, the waxes tend to form clotted accumulations in the depressions, where cells meet; these accumulations are readily evident in Figs 10 and 47. Wax rods that do not merge into amorphous clots and that retain their characteristic form occur in the depressions between cells in the abaxial epidermis of Lonchostoma myrloides (Fig. 24). Where waxes are less abundant, the rods or flakes are sparse, as in Figs 3, 4, 5, 13 and 34, they can be identified by comparison with species in which waxes are more abundant, but some of the objects on such nearly-smooth surfaces may be contaminants. LEAK ANATOMY OF BRUN1ACEAE 17

Figures 35-38. Leaf anatomy of Mtbelia. Figs 35-36. JV. lams i Rourke 709), leaf transections. Fig. 35. Transection of half of leaf, adaxial face above (dark-staining deposits absent because of NaOH treatment;. Fig. 36. Stoma, showing dome of cuticle overarching guard cells. Figs 37-38. A', sphaerocephala [Carlquist 46I4\. SEM photographs of abaxial leaf epidermis. Fig. 37. Trichomes on leaf surface. Fig. 38. Dome-shaped cells with cuticular relief. Scale bars: Fig. 35 = 21 nm, Fig. 36 = 76 urn. Fig. 37 = 10 urn and Fig. 38 = 67 urn. 18 S. CARLQUIST Epidermal cell shape Metcalfe & Chalk (1950) stated that epidermal cells in leaves of Bruniaceae "may be arched outward". This description, apt in terms of sections, can be translated into 'dome-shaped' if a three-dimensional view, such as one obtains with SEM, is available. In fact, epidermal cells in some Brunia leaves are markedly globose in shape (Figs 9, 10, 16). The rounded cells on the three ridges of Linconia alopecuroidea (Figs 17, 18) are similar. The term 'papillate' is used to describe the outer wall of epidermal cells in some Bruniaceae, and this term appears quite apt for cells such as those of the abaxial epidermis of Lonchostoma monogyna (Fig. 18) and L. myrtoides (Figs 24, 25) as well as cells near the leaf-tip of Linconia cuspidata (Fig. 19) or Raspalia stokoei (Fig. 43) or R. virgata (Fig. 45). What appear to be dome-shaped or papillate cells at leaf margins are in some instances short unicellular trichomes (Fig. 5). Nebelia sphaerocephala has markedly papillate or dome-shaped epidermal cells; trichomes are intermixed with these cells (Fig. 37). Dome-shaped cells are illustrated for Staavia radiata (Fig. 47). Only moderate outward bowing of epidermal cell walls is seen in most Bruniaceae (e.g. Figs 4, 7, 40). The shape of epidermal cells as seen from above is chiefly polygonal in outline in Bruniaceae (Figs 24, 46). Exceptions to this occur in the adaxial epidermis of Lonchostoma (Fig. 28) and Raspalia. In these leaves, which are 'inversely' bifacial, epidermal cells with undulate outlines occur on the adaxial surface intermixed with stomata. Leaves in these genera also have a marked difference between adaxial and abaxial epidermis in cell size (Figs 6, 8, 11, 35, 52). Audouinia (Fig. 1) has an adaxial epidermis with cells much smaller than those of the abaxial epidermis, despite the fact that by some criteria (stomatal distribution) this leaf is transitional to isolateral.

Subsidiary cells Photographs of stoma-bearing epidermis from paradermal sections of leaves of Bruniaceae were submitted to H. Rasmussen, who has authored a monograph of stomatal nomenclature and ontogeny (Rasmussen, 1981). Rasmussen has kindly commented on these photographs, and finds that a photograph of Berzelia cordifolia epidermis "is a typical case of cyclocytic" stomata as defined by Stace (1965). The stomatal apparatus illustrated here for Pseudobaeckea africana (Fig. 39) is like that of Berzelia cordifolia. Rasmussen finds that the stomatal apparatus of Lonchostoma pentandrum (Fig. 28), although it shows a less consistent pattern, may also be referred to the cyclocytic concept. The subsidiary cells are formed after the formation of a stomatal meristemoid, by means of mitoses in surrounding protodermal cells, and thus are perigene in origin. Failure of division of some of the protodermal cells surrounding stomata in Lonchostoma would produce the "less consistent" pattern cited for Fig. 28 according to Rasmussen.

Guard cells Guard cells are relatively uniform in Bruniaceae. In terms of external appearance. The oval, shield-shaped appearance of the guard cells as seen in LEAF ANATOMY OF BRUNIACEAE [9

Figures 39-42. Leaf anatomy of Pseudobaeckea and Raspalia. Fig. 39. P. africana i Carl'quist 4784), stoma and subsidiary cells from paradermal section. Fig. 40. P. cordala {Carlquist 4928), SEM photograph of adaxial epidermis, showing cuticular relief. Fig. 41. R. globosa (Carlquist 4811 , transection of leaf, adaxial surface at left. Fig. 42. R. pkylicoides (Carlquist 4890), leaf midvein with fibres and druse {below). Scale bars: Fig. 39 = 76 urn. Fig. 40 = 59 um. Fig. 41 = 21 Jim and Fig. 42 = 76 urn. surface view is illustrated in Figs 4, 5, 7, 14, 27, 40, 47, 48 and 49. The stomatal aperture is more elongate in some of these (Fig. 27), smaller and more nearly circular in others (Figs 47, 49). In sectional view, a cuticular dome overlies the 20 S. CARLQUIST pair of guard cells (Figs 36, 53). The dark-staining wall of the guard cells is thinner externally, thicker internally. Sub-adjacent to the pair of guard cells in sectional view are subsidiary cells, which are triangular in shape. The cuticular domes overlying the guard cells may be divided into two categories: those with a simple dome (Fig. 36), and those with a flange around the outer surface of the dome (Fig. 53). The flange appears as a light circle outlining the circular dome seen in face view in SEM photographs (Figs 17, 20). The flange was observed on stomata of Audouinia, both species of Linconia, Pseudobaeckea teres and all four species of Tittmannia. Within the cuticular dome of the guard cell pair is a chamber. A double chamber (outer plus inner) was figured by Niedenzu & Harms (1930) for Staavia dodii leaves. The doubleness of this chamber is due to an inward extension of cuticular material, subdividing the chamber. However, in my material of Staavia dodii, only the most minimal inward bowing of the cuticle could be observed, so that without the figure of Niedenzu & Harms one would be unlikely to notice anything unusual in the stomatal canal in this species. In figures of guard cells as seen in transection in dicotyledons, one commonly sees not only an outer cuticular ridge by which the stoma is closed, but also an inner ridge on the inner surface of the guard cell pair. Such an inner ridge was observed (in addition to the more prominent outer ridge) in Linconia alopecuroidea, Raspalia barnardii and R. phylicoides. An inner cuticular ridge is very likely much more common in Bruniaceae, but could not be observed because very thin, deeply stained sections demonstrate this ridge optimally. Stomata are raised above the epidermal surface on the adaxial epidermis of Raspalia virgata leaves. The stomatal domes are neither raised nor sunken, but approximately flush with the surface on the adaxial surfaces of leaves of Raspalia globosa, R. microcephala and R. palustris.

Trichomes The data of Pillans (1947) on morphology of Bruniaceae includes adequate summaries of general trichome texture and distribution on leaves. Although earlier accounts are vague on this point, all trichomes of Bruniaceae are non- glandular and uniseriate. The trichomes terminate in points (Fig. 3, centre; Figs 13-15, 37, 50, 53). Some trichomes lose the tip of the terminal cell, and both in sectional view and with SEM the truncated nature of such trichomes may be seen (Fig. 3, left; Fig. 5, right). Trichomes may be lost as leaves mature. Pillans (1947) mentions trichomes only on young leaves of Brunia albiflora, B. alopecuroides and B. nodiflora. Hair remnants may not be evident in these species (e.g. Fig. 9) because trichome remnants have shrivelled and been covered by cuticle. Very likely, as with other angiosperm groups, the difference between glabrous and hairy leaves is one of preservation of trichomes, which are probably present on young leaf primordia in all species of the family. In the genus Brunia, Pillans (1947) mentions hairy covering of adult leaves only for B. laevis and B. macrocephala. Leaves of these two species prove to have trichomes highly distinctive for the family. In B. laevis, large straight trichomes are intermixed with numerous smaller helically coiled trichomes (Fig. 13). In B. macrocephala, trichomes are small and tightly coiled on the abaxial surface of leaves (Fig. 14). On the adaxial surface (Fig. 15), in addition to small coiled I.I.AI W.YIOMY OF BRUMAGEAE 21

Figures 43-46. Leaf anatomy of Raspaiia. Fig. 43. R. stokoei {Carlquist 5053), SEM photograph of abaxial epidermis. Fig. 44. R. variabilis [Carlquist 4966), SEM photograph of trichome on abaxial epidermis. Figs 45-46. R. virgata (Carlquist 4936";, SEM photographs of abaxial leaf surface. Fig. 45. Leaf tip, showing apicula; cells of apicula have broken surfaces, exposing globular deposits within. Fig. 46. Epidermal cells, showing waxy rods on surface. Scale bars: Fig. 43 = 72 um. Fig. 44 = 243 um. Fig. 45 = 15 um and Fig. 46 = 7.3 um. 22 S. CARLQUIST

Figures 47-50. Leaf anatomy of Staavia, Thamnea and Tittmannia. Fig. 47. S'. radiala Carlquist 4949), SEM photograph of epidermal cells and wax deposits on leaf margin. Figs 48-49. Thamnea thesioidts [Esterhuysen 35105), SEM photographs of abaxial leaf surface. Fig. 48. Leaf portion showing apicula and keeled nature of surface. Fig. 49. Two stomata and waxy deposits on leaf epidermis. Fig. 50. Tittmannia esterhuyseniae (Carlquist 5040). SEM photograph of adaxial surfaces of leaves. Scale bars: Fig. 47 = 69 urn, Fig. 48 = 15 Mm, Fig. 49 = 94 ^m and Fig. 50 = 6.2 urn. trichomes there are a few much larger trichomes that may be somewhat curved or straight. Curled trichomes were also observed on the adaxial surfaces of leaves of Raspalia globosa (Fig. 41) and R. virgata (Fig. 45). The two species of Mniothamnea differ with regard to trichome presence: trichomes are abundant on the adaxial surfaces of leaves of M. bullata (Fig. 31). LEAF ANATOMY OF BRUNIACEAE •:.;

Figures 51-54. Leaf anatomy of Tillmannia laxa [Carlquist 5051). Fig. 51. Section of apicula from leaf longisection, showing mild phellogen activity and dark-staining cell contents. Fig. 52. Leaf transection, abaxial face below. Fig. 53. Stoma and trichome from leaf transection. Fig. 54. Palisade thlorcnchyma (above, fibres on midvein i below i and rhomboidal crystal from leaf transection. Scale bars: Fig. 51 = 76 am, Fig. 52 = 21 um. Fig. 53 = 76 urn and Fig. 54 = 76 urn.

They are sparse on leaves of M. callunoides (Fig. 34), in which they occur chiefly on margins and leaftips. Nebelia sphaerocephala (Fig. 37) differs from the other species of JVebelia in the density of hairs on its leaves. Leaves of Staavia are 24 S. CARLQUIST virtually hairless at maturity. In Tittmannia esterhuyseniae, the leaves have elongate trichomes on the margins, whereas in the other species of Tittmannia, which have terete leaves, the trichomes are very short and on all leaf surfaces (Figs 52, 53). The two species of Linconia are quite different on the basis of trichomes: L. alopecuroidea (Fig. 17) has slender trichomes borne densely on the margins, whereas trichomes are lacking on leaves of L. cuspidata (Fig. 20). Small pointed trichomes of Bruniaceae, like those of Audouinia (Fig. 3), Berzelia (Fig. 5, right) and Tittmannia (Fig. 53) are cutinized, and stain faintly with safranin but not at all with fast green. However, the long trichomes of Linconia alopecuroidea, Lonchostoma (all species), Nebelia jragarioides, Raspalia barnardii, R. dregeana, R. microphylla, R. palustris, R. phylicoides, R. sacculata, R. variabilis, R. villosa and R. virgata are little cutinized, and stain prominently with fast green. In these species, higher power SEM observation reveals longitudinal striations on the surface. These striations, figured here for Raspalia variabilis (Fig. 44), are equally prominent in Lonchostoma, and may even be seen under high power with the light microscope.

Contents of mesophyll and epidermal cells In palisade and spongy parenchyma of leaves of Bruniaceae one can see various degrees of presence of two kinds of compounds. One of these tends to fill cells with massive brightly staining deposits, often globular in shape. These deposits are yellowish in the unstained specimen and stain bright red with safranin. They may be terpenoids, but identification is needed. The other common material encountered is granular and greyish; it stains dark blue to greyish in colour, and is tentatively identified as a tannin here. Both of these materials can be present within a single cell. The massive reddish-staining deposits tend to fill all of the palisade cells, but to be present in only part of the spongy mesophyll cells. This mode of occurrence is cited as'S' in Table 1, column 11; presence of these deposits in a portion of the palisade cells is denoted as V. Presence of these deposits in all palisade cells is illustrated here for Audouinia capitala (Fig. 1), Berzelia cordifolia (Fig. 6), Lonchostoma esterhuyseniae (Fig. 22), L. purpureum (Fig. 29), Mniolhamnea bullata (Fig. 32), Raspalia globosa (Fig. 41) and Tittmannia laxa (Figs 52-54). Reddish-staining deposits characteristically occur in the leaf apiculae (Fig. 51). The more limited presence of reddish-staining compounds (V in Table 1, column 11) is illustrated here for Berzelia squarrosa (Fig. 8), Brunia alopecuroides (Fig. 11), and Linconia cuspidata (Fig. 21). Epidermal cells only occasionally contain the reddish-straining deposits, e.g. as illustrated here for Lonchostoma esterhuyseniae (Fig. 23). Epidermal cells do often contain the granular substance identified as tannins here. Tannins in epidermal cells can be seen in Fig. 29 (lower left) and Fig. 52 (especially lower right). Epidermal cells adjacent to guard cells often contain tannins and thereby stain more darkly (Fig. 28). Tannins, where they tend to occur in all or nearly all mesophyll cells, are designated by T in Table 1, column 11. In species in which tannins are less abundant, some tannin-free spongy mesophyll cells may be seen (V in Table 1, column 11). Tannins are notably abundant in leaves of Nebelia. Palisade cells may contain both the reddish-staining materials and the tannins, and thus LEAF ANATOMY OF BRUNIACEAK 25 designation of the mode of occurrence of either one is necessarily imprecise. Where tannins are abundant, as in Nebelia, all mesophyll cells stain dark reddish-purple. Dried materials have been utilized for some species. The NaOH used to treat these specimens preparatory to sectioning tends to dissolve the reddish-staining compounds and tannins to various extents. Estimation of the presence of mesophyll contents in these species is therefore subject to error.

Crystals Two types of calcium oxalate crystals are present in leaves of Bruniaceae, and these two types of crystals are also associated, with few exceptions, with distinctive associated modes of fibre occurrence. Rhomboidal crystals occur singly in bundle sheath cells; they can be seen conspicuously in the bundle sheath cells immediately outside fibres adjacent to phloem of veins, especially the midvein (all veins in leaves of some species, such as Thamnea diosmoides). The rhomboidal crystals occur in bundle sheath cells only in genera in which massive fibre strands occur in the leaf mid-vein. As shown in Table 1, column 12, this mode of crystal occurrence ('R') characterizes the genera Audouinia, Berzelia, Brunia, Nebelia, Thamnea and Tittmannia. Examples are illustrated here for Audouinia capitata (Fig. 2, top), Brunia alopecuroides (Fig. 12, upper left and lower left) and Tittmannia laxa (Fig. 54, centre right). In the alternative mode of crystal occurrence, druses occur not in bundle sheath cells but in mesophyll cells, chiefly in spongy parenchyma. As indicated in Table 1, column 123, this mode of crystal occurrence ('D') characterizes the genera Linconia, Lonchosloma, Mniolhamnea, Pseudobaeckea, Raspalia and Staavia. Examples are illustrated here for Lonchosloma esterhuyseniae (Figs 22, 23), L. purpureum (Figs 29, 30), Raspalia globosa (Fig. 41), and R. phylicoides (Fig. 42). Niedenzu & Harms (1930) and Metcalfe & Chalk (1950) reported absence of crystals in Raspalia; these reports must be disregarded. Crystals were not observed in two species of Raspalia studied, R. sacculata and R. virgata, but exhaustive search of material of these species might reveal presence of druses. Pseudobaeckea teres does not agree with other species of Pseudobaeckea in crystal occurrence; there are various reasons to believe that this species is misplaced. Although the dichotomy in the family with respect to modes of crystal occurrence is nearly perfect, a few species in the first category have variations that do not really violate the difference outlined but do show some diversity. In leaves of Audouinia capitata and Berzelia rubra, a few rhomboidal crystals were observed in mesophyll cells in addition to those in the bundle sheath. In Berzelia rubra, Nebelia sphaerocephala and Thamnea diosmoides, rhomboidal crystals are present in bundle sheath cells as specified for those three genera, but in addition a few druses occur in mesophyll cells.

Veins In Table 1, column 13, the number of veins per leaf (as seen in a typical transection) is recorded. The number of bundles per leaf undoubtedly varies greatly, and the numbers given in Table 1 cannot represent variation within a species because of the small number of leaves sectioned. The values given are 26 S. CARLQUIST representative of whether one, a few, or many veins are present, however. Metcalfe & Chalk (1950) recognized a range in number of veins, but considered three a common number. In my experience, the range in vein number is almost continuous, related to leaf width, and three is not notably common. When dried leaves are viewed externally, one can sometimes detect veins as eminences, but the thickness of leaves and cuticle renders such vein indications imprecise. In wider leaves, such as those of Berzelia cordifolia and Lonchostoma myrtoides, one gains the impression from viewing leaf surfaces that the veins must branch dichotomously and terminate freely. In fact, paradermal sections show that this is not true. Marginal veins were regularly observed in paradermal sections, and few free vein terminations (no more so than in the average dicotyledonous leaf) could be observed. Certainly anastomoses of veins are common. No teeth are present on margins of bruniaceous leaves, so lack of vein termini along the margins is not surprising. Smaller veins may lack phloem. Although veins devoid of phloem may occur anywhere within a lamina, they are slightly more common toward the leaf margins, perhaps because veins are smaller there. As viewed in three dimensions, the veins in leaves of Bruniaceae form a planar or two-dimensional system with the exception of the genus Staavia. In Staavia, a secondary set of veins lies abaxial to the main set of veins in all species. These veins are not easily illustrated because they are small and embedded in darkly- staining mesophyll. All of the secondary veins are 'normal' in orientation, that is, with xylem adaxial to phloem. One secondary vein was seen in leaves of Staavia capilella, two in S. glutinosa and S. verticillata, three in S. zeyheri, four in S. pinifolia, and six in the broad-leaved species, S. dodii. The secondary veins are included in the numbers given in Table 1, column 13. All tracheids, except for storage tracheids in leaves of Bruniaceae, have annular (less commonly) or helical (commonly) thickenings. As Niedenzu & Harms (1930) noted, Marloth's (1925) figures showing only pitted tracheids in leaves of the family are in error. However, they are only partly in error, because storage tracheids, which Marloth illustrated at the tip of a leaf of Linconia alopecuroidea, prove to be pitted in Bruniaceae. Fine helical sculpture facing the lumen can also be observed in storage tracheids. The term 'storage tracheid' is applied here to tracheids that are relatively isodiametric in shape, and are wider than ordinary tracheids. Storage tracheids are characteristically found in the vein terminus that underlies the leaf apicula throughout the family. They also occur in some species at the leaf margins, most notably in Staavia dodii, in which they are also notably common in the secondary veins. In Table 1, column 14, presence and relative abundance of fibre strands in leaf veins is given. A large strand of fibres on the midvein, but on few other veins on the leaf, is characteristic (as mentioned above in connection with modes of crystal occurrence) of Audouinia (Figs 1, 2), Berzelia (Fig. 6, right), Brunia (Figs 11, 12), Nebelia (Fig. 35, dark area at right), Thamnea and Tittmannia (Figs 52, 54). Small strands of fibres, present on the midvein and rarely also in some of the lesser veins (e.g. Lonchostoma myrtoides) are present in the genera that have druses in mesophyll but lack rhomboidal crystals in bundle sheath cells. These smaller fibre strands are illustrated for Lonchostoma (Figs 22, 29) and Raspalia (Fig. 41). This condition was also observed in Mniothamnea, Pseudohaeckea (except P. teres, as LEAF ANATOMY OF BRUNIACEAE 27 noted above, which has a massive strand of fibres on the mid vein), and Staavia. No fibres were observed on veins in leaves of Linconia (Fig. 21), Mniothamnea caltunoides, and some collections of Pseudobaeckea and Raspalia (see Table 1, column 14). The leaf of Nebelia sphaerocephala is anomalous in having at each margin a large strand of fibres that is not associated with any vein. The midvein fibres of Audouinia do not stain red with safranin, as do those of other Bruniaceae. The walls stain bluish with fast green. More significantly, shrinkage spaces are evident in the fibre walls. These cells can therefore be called gelatinous fibres.

ECOLOGICAL CONCLUSIONS One can construct quantitative formulae incorporating data from leaves in order to arrive at numerical indices of degrees of xeromorphy or mesomorphy. Such a formula has been used in order to compare leaves of Hawaiian species of Geranium with each other (Carlquist & Bissing, 1976). That formula seemed a sensitive indicator of degrees of mesomorphy in that genus, but the comment was made in that paper, "it is a strictly arbitrary index, and would not necessarily be applicable to other groups of ". The formula in Carlquist & Bissing I 1976) incorporated number of leaf teeth and leaf thickness. These features are not applicable to Bruniaceae. There is only a single tooth in leaves of Bruniaceae (if one can term the apicula a tooth at all). Small, scale-like leaves of Bruniaceae are doubtless indicative of xeromorphy, but because of the small size of these leaves, their limited thickness would make them appear much more mesomorphic (which they are not) than acicular leaves that arc thicker. If these two dimensions are omitted from the formula used by Carlquist & Bissing (1976), one obtains the formula:

where A = leaf area, £", = cuticle thickness of adaxial epidermis and E.2 = cuticle thickness of abaxial epidermis. Area has been computed for species of Bruniaceae (Table 1, column 15) by using the formula of length times width divided by two, a formula based on the idea that leaves are variously elliptical or triangular in outline, and that this formula applies to this range of shapes. Area is given for leaves of the species studied in Table 1, column 15. The mesomorphy value (M) for leaves given by the formula above is expressed for the species studied in Table 1, column 16. The values obtained by the leaf mesomorphy formula range from 0.01 (Thamnea thesioides) to 2.87 {Lonchostoma myrtoides). If the formula given in the preceding paragraph were applied to leaves of the Hawaiian species of Geranium, the values range from 17.9 {Geranium tridens Hillebrand) to 3929 (Geranium arboreum Gray). Even though the Hawaiian species of Geranium as a whole occupy dry sites, their leaves are more mesomorphic than those of the most mesomorphic of Bruniaceae. If the formula in the preceding paragraph were applied to rain forest trees, or understory plants, values much higher than those for the Hawaiian Geranium species would be obtained. Thus, leaves of Bruniaceae are extraordinarily xeromorphic by comparison with those of angiosperms at large. Woods of Bruniaceae are xeromorphic but not strongly so compared with those 28 S. CARLQUIST of dicotyledons at large (Carlquist, 1978a). This leads one to the conclusion that leaf xeromorphy may be of primary importance, wood structure much less important, in governing the water economy of Bruniaceae. Although Bruniaceae can often be found in moist microclimatic sites, or sites with protracted moisture availability (seeps), the low humidity and frequent strong winds of Cape Province provide dry aerial conditions that make xeromorphic leaves adaptive. If one reviews the leaf mesomorphy values (Table 1, column 16), one notes a considerable range: the 2.87 of Lonchostoma myurtoides is approximately 300 times the 0.01 of Thamnea thesioides. If one compares the values for the various species with their known habitats, one finds valid correlations (locality data from Pillans, 1947, and original observations). Thamnea thesioides occurs on the lower » slopes of mountains near Ceres, which lies close to the Karoo Desert. Similar correlations can be found for other species with leaf mesomorphy values below 0.10. Most of the species with values below 0.10 are from relatively dry montane localities. Bruniaceae with notably high leaf mesomorphy values tend to be in moister habitats. Lonchostoma myrtoides grows in wet acid sand flats, the sort of localities where one finds Drosera. These localities are termed 'vlakte' or sometimes 'swamps'. The same kind of habitat is occupied by the species with the next highest leaf mesomorphy value, Lonchostoma pentandrum. Lonchostoma monogynum (0.58) also grows on these wet flats, but L. purpureum (0.56) occurs in montane meadows. Lonchostoma esterhuyseniae, which has the lowest mesomorphy value in its genus, grows in moderately dry montane areas, but on south-facing slopes and in the shade of boulders. Pseudobaeckea cordata has a relatively high leaf mesomorphy value; it frequents wet places such as meadows, swamps and wet stream-banks. In other genera, a high value occurs in Berzelia cordifolia (1.32). This species occupies lowland hills that seem dry. However, the leaves of B. cordifolia are almost the thickest in the family (Table 1, column 3), and this thickness, not taken into account in the leaf mesomorphy formula, is undoubtedly of compensatory significance. Mniothamnea offers an interesting comparison. Of the two species, the higher leaf mesomorphy value (0.20) occurs in M. bullata, which grows in moist shady crevices of the Clock Peaks. Mniothamnea callunoides (0.09) occurs in fully exposed sites lower in the Clock Peaks. Leaf area within the family has probably been subject to reversions. The varying leaf forms within a single species, Pseudobaeckea cordata, suggest this. Wider leaf forms such as those of Lonchostoma myrtoides and Berzelia cordifolia are likely to be secondary, based on the fact that other species in these genera have narrower leaf forms, and neither of these genera is considered to have a preponderance of features judged to be primitive for the family (e.g. Pillans, 1947). Although leaf area and cuticle thickness are of primary importance in restricting transpiration, the next most important feature of bruniaceous leaves with respect to xeromorphy is one that is difficult to quantify: the tendency of leaves to be appressed to stem surfaces. The genera in which this feature is most evident are Mniothamnea (Figs 31, 34), Raspalia, Thamnea (T. thesioides, Fig. 48), and Tittmannia. The effectiveness of this strategy may be suggested by the fact that these genera have very low leaf mesomorphy values: most species in these genera fall below 0.20. LEAF ANATOMY OF BRUNIACEAE 29 Despite the tendency of leaves to be appressed to stem surfaces, stomata in Thamnea occur on the abaxial face (Fig. 48). In fact, the majority of species with appressed leaves do have stomata on both leaf surfaces (Table 1, columns 9 and 10). In Raspalia, the adaxial leaf surface, closely appressed to the stem, is reminiscent of the infolded surface of an ericoid leaf in some respects: the adaxial surfaces of/?, globosa leaves bear hairs and raised stomata (Fig. 41, left), features of ericoid leaves such as those of Empetrum (Metcalfe & Chalk, 1950). The tendency of leaves to be appressed to stems is shown to a moderate extent in Audouinia, Lonchostoma, Nebelia, and, to various degrees, Berzelia and Brunia. In species with leaves appressed to stem surfaces, the leaves tend to be inversely bifacial rather than isolateral an adaptation one might expect because so little light reaches the adaxial surface. In species in which leaves are less appressed or not at all appressed to stems (Berzelia pro parte, Brunia pro parte, Linconia, Staavia). one finds isolateral leaf construction. In these species, restriction of leaf surface by means of an acicular shape is a prominent way of minimizing transpiration. The two collections of Pseudobaeckea cordata form an interesting comparison. The collection Carlquist 4716 has thinner cuticle (and less cuticular relief) on the abaxial surface, and more nearly normally bifacial leaves (stomata are more abundant on the abaxial surface); the collection Carlquist 4928 has thinner cuticle (and less cuticular relief) on the adaxial surface, and shows a closer approach to isolateral leaf construction (stomata are about equally frequent on both surfaces). The above comparison is noteworthy in that the features are not at random, but are positively correlated with each other. Moreover, they show the tendency for cuticular relief to be more pronounced on the leaf surfaces more exposed to the environment. Cuticular relief in Bruniaceae may have the effect of aiding radiation of heat from the leaves. The prominent cuticular dome that overlies guard cells in Bruniaceae (e.g. Figs 4, 5, 7, 27) must be counted a xeromorphic feature. The prominent overarching by cuticle provides a pocket that potentially attenuates the humidity gradient between the leaf interior and the environment. This can be considered a formation analogous to the pockets ('stomatal crypts') in leaves of Banksia, Ceanolhus and JVerium (Metcalfe & Chalk, 1950).

SYSTEMATIC CONCLUSIONS Although Pillans (1947) offered a phylogenetic tree to the genera, he proposed no system of tribes or subfamilies. However, a system was offered by Niedenzu & Harms (1930), who divided genera between tribe Audouinicac (Audouinia. Thamnea, Tittmannia) and Brunieae (the remaining genera) based on stamen morphology. Their distinction seems difficult to define clearly, because several genera appear intermediate in stamen morphology on the basis of drawings by Pillans (1947). Stipules are present in Brunia, Berzelia, Linconia, Staavia and Tittmannia according to Dahlgren & Van VVyk (1988), but that grouping of genera does not seem to form a natural subfamily. The distribution of stipules may not be as discrete as the listing indicates; possibly all genera of Bruniaceae form stipule primordia ontogeneticallv, but some mature these into visible stipules while others do not. According to Pillans (1947), some species of Berzelia have stipules, whereas others lack them. 30 S. CARLQUIST Goldblatt (1981) demonstrates three levels of ploidy in the family: n = 11 in Audouinia; n = 22 (or close to that) in Brunia, Lonchostoma, Nebelia, Raspalia and Slaavia; and n = c. 40 (perhaps n = 44, since accurate counting was impossible) in Berzelia. The basic chromosome number appears to confirm the naturalness of the family, but it does not help with intrafamilial taxonomy. The basic position within the family accorded to Audouinia by various authors is supported by chromosome data. Data from pollen morphology of Bruniaceae are much like those for cytology in that a tricolpate form occurs in the genera Audouinia, Berzelia, Mniolhamnea and Pseudobaeckea, but four or more colpi characterize pollen of the remaining genera (Hall, 1988). Species of the genus Brunia fall into two groups on this basis, and Hall (1988) uses this division to suggest transfer of three species of Brunia to Berzelia. Data from pollen alone do not offer good bases for subfamilial classification, although they do clearly show a phyletic trend away from a basic tricolpate type. Takhtajan (1987) modifies the Niedenzu & Harms (1930) system of tribes. He recognizes Audouinioideae (Audouinia, Thamnea, Tiltmannia), Brunioideae (Linconia, Raspalia, Nebelia, Slaavia, Pseudobaeckea, Brunia), Lonchostomoideae {Lonchostoma) and Berzelioideae (Mniolhamnea, Berzelia). Presumably Lonchostoma is segregated on the basis of its sympetalous corolla. The close affinity between Berzelia and Brunia inferred by Pillans (1947), perhaps because of the capitate inflorescences shared by these genera, is not evident in the Takhtajan treatment. Mniolhamnea and Berzelia share a unilocular ovary containing a single ovule, and doubtless Takhtajan is emphasizing that. Certainly the unilocular ovary seems a derived feature. The species of Brunia are transitional to that condition according to Pillans (1947). Audouinia is the only genus of the family with three carpels. The remaining genera have two carpels or one, variously separate from each other and from other floral parts. Species of Bruniaceae have a range from almost completely superior ovaries to almost completely inferior ones, but this feature varies within genera and does not seem a feature of significance for subfamilial classification. The basal position within the family of Audouinia, claimed by all authors, is supported by data from cytology (Goldblatt, 1981) and pollen (Hall, 1988) as noted above. In addition to the tricarpellate nature of the ovary, Audouinia is probably judged primitive because flowers are borne on short shoots. Short shoots of a reduced nature can be said to be present in other genera, but in the form of bracteolate pedicels: Thamnea has numerous bracteoles, and 5-8 are present beneath each flower in Tiltmannia. Linconia flowers are subtended by 4—6 bracteoles, there are fewer in the other genera and only a single one in Brunia. The vestiges of the short shoot in Audouinia, Thamnea and Tiltmannia can be cited in support of their inclusion within Audouinioideae and in support of the primitiveness of that group. In view of the varied nature of evidence on systematics and phylogeny within the family, one naturally asks what kinds of evidence might be offered by leaf anatomy. One can, in fact, express the data from leaf anatomy in the form of a key: Large strand of fibres adjacent to phloem of midvein; rhomboidal crystals present in bundle sheath cells of midvein Fibres of midvein gelatinous Audouinia LEAF ANATOMY OF BRUNTACEAE 31 Fibres of midvein lignified Leaf mesophyll very densely staining, both palisade and mesophyll cells filled with tannins and amorphous deposits Nebelia Leaf mesophyll less densely staining, at least some spongy- cells devoid of tannins or amorphous deposits Stomata present only on abaxial surface, which has a cuticle thicker than adaxial surface Thamnea Stomata present on both surfaces, or if lacking on the adaxial, adaxial cuticle less than 9 um thick Leaves nearly terete, usually less than twice as wide as thick Brunia, Tittmannia pro parte Leaves flatter, less than twice as wide as thick Berzelia, Tittmannia esterhuyseniae Fibre strands associated with veins composed of few fibres or fibres lacking on veins; druses present in mesophyll Stomata present on both surfaces More than one set of bundles present, the secondary set abaxial to the main one but with normal (xylem adaxial) orientation; fibres present on the main veins Staavia Only a single set of bundles present; fibres absent on main veins or very few Pseudobaeckea Stomata present on adaxial surface only Adaxial cuticle 2 |im or thinner; leaves less than 2 mm wide Mniothamnea Adaxial cuticle more than 2 p.m thick; leaves more than 2 mm wide Lonchostoma Stomata present on abaxial surface only Leaves more than twice as wide as thick Raspalia Leaves less than twice as wide as thick Leaves triquetrous, with long hairs, more than 10 mm long Linconia alopecuroidea Leaves oval in transection, glabrous, less than 10 mm long Linconia cuspidata It should be noted that Pseudobaeckea teres (assuming that the specimen studied has been correctly determined) does not conform to the above key—it falls into the first part of the key (in which it would key out together with Tittmannia esterhuyseniae), whereas Pseudobaeckea without P. teres falls in the second half of the key. Pillans (1947) doubted whether P. teres belongs in Pseudobaeckea, and suggested Raspalia as the genus in which it should be located. However, Raspalia lacks the large group of fibres on the midvein and the rhomboidal crystals on the bundle sheath whereas Pseudobaeckea teres has these distinctive features. Pseudobaeckea teres appears closer to Tittmannia than to any other genus. Pollen data offered by Hall (1988) would also tend to ally Pseudobaeckea teres with one of the genera with three colpi (e.g. Audouinia, Thamnea, or Tittmannia), although the exine texture of pollen of P. teres is distinctive. At the end of the above key, the two species of Linconia have been keyed separated. Although Linconia may well be a natural genus in terms of floral features, the two species are highly distinctive, and have relatively few leaf 32 S. CARLQUIST features by which the two together may be contrasted with the other elements in the second half of the key. Assuming that Pseudobaeckea teres can be placed more satisfactorily, one may- ask whether the key given above corresponds to subfamilial groupings. The unusual combination of a character involving a large strand of fibres with a character involving occurrence of rhomboidal crystals in bundle sheath cells suggests that this may be a natural subdivision of the family. Molecular studies are needed to see whether groupings in the above key are natural or not. Wood of Bruniaceae either lacks crystals or, if it has them, they are rhomboidal crystals only, with druses rudimentary at most (Carlquist, 1978a). This suggests that the genera that have rhomboidal crystals in the leaves may have retained a more r primitive feature. This circumstance would agree with other types of evidence (Pillans, 1947; Goldblatt, 1981; Hall, 1988) that suggest a more primitive position for Audouinia, along with Thamnea and Tittmannia. *

PHYLOGENETIC CONCLUSIONS Dahlgren & Van Wyk (1988) have offered a summary of relationships of Bruniaceae. Their concept is that Bruniceae and Grubbiaceae are closely related, a concept originated by Van Tieghem (1897). Grubbiaceae has been claimed by some authors (e.g. Fageriind, 1947; Cronquist, 1988) to be referable to Ericales. Therefore, some authors, such as Cronquist, have placed Grubbiaceae in Ericales, and Dahlgren & Van Wyk (1988) also claim an ericalean affinity for Bruniaceae and Grubbiaceae, although they concede there is little evidence on this point. The alternative view is that Bruniaceae are rosalean or even hamamelidalean. Viewpoints of this nature have been endorsed by Van Tieghem (1897), Xicdenzu & Harms (1930), Thome (1976), Takhtajan (1987); a review of this position is given by Niedenzu & Harms (1930). Thome's treatment (Thome, 1976) is of interest, because he groups several South African families (Bruniaceae, Geissolomataceae, Grubbiaceae, and Myrothamnaceae) in the suborder Brunineae of Pittosporales. Thorne considers that Pittosporales might be allied closely with either or Hamamelidales (personal communication). With respect to leaf anatomy, there are some reasons to ally Bruniaceae with Grubbiaceae. Leaves of both families are linear, without teeth. The mesophyll cells contain reddish-staining massive deposits. Trichomes are thick-walled and non-cutinized, like those of Lonchostoma or Raspalia, and trichomes in both • families are uniseriate and exclusively non-glandular. Grubbiaceae have rhomboidal crystals in bundle sheath cells as in six genera of Bruniaceae, but they have druses in mesophyll similar to the other six genera (Carlquist, 1978b); they also have rhomboidal crystals in mesophyll (as in Audouinia). The evidence from trichomes is interesting, because basic trichome types are rather more conservative than is generally appreciated. The families of Ericales have glandular trichomes (Metcalfe & Chalk, 1950), and thereby differ from Bruniaceae and Grubbiaceae. Dahlgren & Van Wyk (1988) also cite unitegmic tenuinucellate ovules as allying Bruniaceae with Grubbiaceae, although they suggest that the ovules may not be truly tenuinucellate. Conceding that the Bruniaceae and Grubbiaceae are related to each other, to LEAF ANATOMY OF BRUNIACEAE 33 what other families should one ally them? Dahlgren & Rao (1969), in a study on Geissolomataceae, showed that Geissoloma has crassinucellate bitegmic ovules, thereby forming a potential contrast with Bruniaceae and Grubbiaceae in the opinion of Dahlgren & Van Wyk (1988). As noted above, the ovules of Bruniaceae and Grubbiaceae may not be truly tenuinucellate, so that only the integument number is left as a source of contrast. Despite that feature, Thorne (1976) and Takhtajan (1987) place Geissolomataceae near Bruniaceae. Cronquist (1988) places Geissolomataceae in Celastrales, although his commentary indicates that he has reservations about that placement. These treatments show that even if Grubbiaceae is the family closest to Grubbiaceae, various workers have difficulty citing any other families with affinity to Bruniaceae. The alliances that form Hamamelidales and Rosales are not known in sufficient detail to permit one to place Bruniaceae within a familial sequence with any reasonable degree of conviction. Niedenzu & Harms (1930), although they opted for a rosalean affinity of Bruniaceae, felt that it is isolated within Rosales. Chemical evidence by Jay (1975) and evidence on waxes by Fehrenback & Barthlott (1988) add more reasons for a rosalean position of Bruniaceae, but do not clarify which families might be close neighbours of Bruniaceae.

ACKNOWLEDGEMENTS Elsie Esterhuysen accompanied me on numerous field trips in Cape Province and showed me localities for Bruniaceae. Because of her effort and invaluable knowledge, this paper is dedicated to her. John Rourke kindly placed the facilities of the Compton Herbarium at my disposal and thereby greatly aided my research. Thanks are due David C. Michener for his microtechnical work on leaves of many of the species.

REFERENCES

BOND, P. & GOLDBLATT, P., 1984. Plants of the Cape flora. A descriptive catalogue. Journal of South African Botany, Supplement 13. CARLQUIST, S., 1977. Wood anatomy of Grubbiaceae. Journal of South African Botany, 43: 129-144. CARLQUIST, S., 1978a. Wood anatomy of Bruniaceae. Aliso, 9: 323-364. CARLQUIST, S., 1978b. Vegetative anatomy and systematics of Grubbiaceae. Bolaniska Notiser, 131: 117-126. CARLQUIST, S. & BISSING, D. R., 1976. Leaf anatomy of Hawaiian geraniums in relation to ecology and taxonomy. Biotropica, 8: 248-259. COLOZZA, L., 1904. Bruniacee degli erbari fiorentini. Annali di Botanica Romana. 2: 1-30. CRONQUIST, A., 1988. The Evolution and Classification of Flowering Plants. 2nd edition. Bronx, New York: The New York Botanical Garden. DAHLGREN, R. M. T., 1980. A revised system of classification of the angiosperms. Botanical Journal of the Linnean Society, 80: 91-124. DAHLGREN. R. M. T. & RAO, V. S., 1969. A study of the family Geissolomataceae. Bolaniska .Voliser, 122: 207-227. DAHLGREN, R. M. T. & VAN WYK, A. E., 1988. Structures and relationships of families endemic to or centered in southern Africa. Monographs in Systematic Botany from the Missouri Botanical Garden, 25: 1-94. FAGERLIND, F. S., 1947. Die systematische Stellung dcr familie Grubbiaceae. Svensk botanisk Tidskrift. 41: 315-320. FEHRENBACK. S. & BARTHLOTT, W., 1988. Mikromorphologie der Epicular-Wachse der Rosales s. 1. und deren systematische Gliederung. Botanisches Jahrbuch Systematik, 109: 407^28. GOLDBLATF, P.. 1981. Chromosome cytology of the Bruniaceae. Annals of the Missouri Botanical Garden, 68: 546-550. 34 S. CARLQUIST HALL, A. V., 1988. Systematic palynology of the Bruniaceae. Botanical Journal of the Linnean Society, 96: 285-2%. HALL, A. V. & VELDHLIS, H. A., 1985. South African Red Data Book: plants—fynbos and Karoo biomes. South African National Scientific Programmes Report. 177: 1-157. JAY, M., 1975. Distribution des flavonoi'des chez les Bruniacees. Taxon, 17: 484-488. KIRCHNER, R., 1904. Beitrdge zur Kenntnis der Bruniaceen. Dissertation, Breslau. MARLOTH, R., 1925. The flora of South Africa, 2. Cape Town: Darter Bros. & Co. METCALFE, C. R. & CHALK, L., 1950. Anatomy of the Dicotyledons. 2 vols. Oxford: Clarendon Press. NIEDENZU, F. & HARMS, H„ 1930. Bruniaceae. In A. Engler & K. Prand (Eds), Die naturlnlun Pflanzenfamilien, Aufl. 2, 18a: 288-303. PILLANS. N. S., 1947. A revision of Bruniaceae. Journal of South African Botany. 13: 121-206. POWRIE, E., 1969. A new species of Tittmannia (Braniaceac). Journal of South African Botany, 35: 363-366. RASML'SSEN, H., 1981. Terminology and classification of stomata and stomatal development—a crilical survey. Botanical Journal of the Linnean Society, 83: 199-212. STAGE, C. A., 1965. Cuticular studies as an aid to taxonomy. Bulletin of the British Museum, Botany, 4: 3-78. STRID, A., 1968. A new species of Lonchostoma Bruniaceae). Botaniska Notiser. 121: 312—316. TAKHTAJAN, A., 1987. Syslema Magnoliophytorum. Leningrad: Officina Editoria Nauka. THORNE. R. F., 1976. A phylogenetic classification of the Angiospermae. Evolutionary Biology, 9: 36-106. VAN TIEGHEM, P., 1897. Sur les caracteres et les afFinites des Grubbiacees. Journal de Botamque (Paris). II: 127-128.

#