Aliso: A Journal of Systematic and Floristic Botany Aliso: A Journal of Systematic and Floristic Botany Wood Anatomy of Caryophyllaceae: Ecological, Habital, Wood Anatomy of Caryophyllaceae: Ecological, Habital, Systematic, and Phylogenetic Implications Systematic, and Phylogenetic Implications

Wood of Caryophyllaceae is more diverse than has been appreciated. Imperforate tracheary elements may be tracheids, fiber-tracheids, or libriform fibers. Rays may be uniseriate only, multiseriate only, or absent. Roots of some species (and sterns of a few of those same genera) have vascular tissue produced by successive cambia. The diversity in wood anatomy character states shows a range from primitive to specialized so great that origin close to one of the more specialized families of Cheno podiales, such as Chenopodiaceae or Amaranthaceae, is unlikely. Caryophyllaceae probably branched from the ordinal clade near the clade's base, as cladistic evidence suggests. Raylessness and abrupt onset of multiseriate rays may indicate woodiness in the family is secondary. Successive cambia might also be a subsidiary indicator of secondary woodiness in Caryophyllaceae (although not necessarily dicotyledons at large). Although a small shrub, Gymnocarpos may be primitively woody, and because that genus appears primitive in many wood features the family as a whole may derive from woody ancestors, despite apparent secondary woodiness in many phylads. Systematic distribution of wood character states in the family corresponds closely to the Pax and Hoffmann system of three tribes and their progressive degrees of specialization (Paronychieae, Alsineae, Sileneae). Wood data support the idea that Caryophyllaceae represent a branch from near the base of the order Caryophyllales. Wood of Caryophyllaeeae is highly xeromorphic, comparable in quantitative vessel features to wood of desert shrubs; insular species have less xeromorphic wood. Instances of storying and druse presence in axial and ray parenchyma are newly reported for the family, as is the inverted orientation of xylem, phloem, and periderm produced by a cambium at the periphery of the pith in Dianthus


INTRODUCTION
Few families of dicotyledons are so poorly known with respect to wood anatomy as Caryophyllaceae. The summary of wood histology of the family offered by Metcalfe and Chalk (1950) occupies only four lines. Quantitative data have been offered for wood of only one species (Fahn et al. 1986). Wood anatomy of Caryophyllaceae is, however, of great potential importance in relation to systematics, ecological wood anatomy, and phyletic change in habit. The lack of studies in wood anatomy of the family may be related to the predominantly herbaceous nature of the family. The woodiest representatives (likely the Hawaiian genus Schiedea) are small shrubs the basal stem diameter of which probably rarely exceeds 3 em.
Caryophyllaceae have been invariably regarded as a member of the order formerly known as Centrospermae and currently termed Caryophyllales (less frequently, Chenopodiales), although Caryophyllaceae lack the betalains (betacyanins), glycoftavones, and some other compounds characteristic of other families in the order (Wohlpart and Mabry 1968;Richardson 1981). Sieve-tube plastid ultrastructure shows similarities between Caryophyllaceae and certain other families of the order-Achatocarpaceae, Molluginaceae, and Stegnospermaceae (Behnke et al. 1983). A cladistic study (Rodman et al. 1984) shows Caryophyllaceae as branching from near the base of the order Chenopodiales, well distanced from the betalain-containing families. A wood study claims affinity between Stegnospermaceae and Caryophyllaceae (Bedell 1980). The DNA evidence presently available, based on rbcL data, places Caryophyllaceae in an advanced position in the order, departing from the clade near Chenopodiaceae (Chase et al. 1993). The present study offers some evidence on the phyletic status of Caryophyllaceae.
Wood anatomy of Caryophyllaceae proves to be of especial significance with relation to habit. No Caryophyllaceae are truly shrubby or arboreal. Habits in the family range from subshrubs to perennials with thickened roots to minimally woody perennials to annuals. Relating the range of habits to the rayless wood predominant in the family as well as to the presence of successive cambia (largely in roots rather than Key to columns: 1 (VG), mean number of vessels per group; 2 (VD), mean lumen diameter of vessels at widest point, J.Lm; 3 (VM), mean number of vessels per mm 2 ; 4 (VL). mean length of vessel elements, J.Lm); 5 (VW), mean wall thickness of vessels, J.Lm; 6 (TL), mean length of imperforate rracheary elements, J.Lm; 7 (TW), mean wall thickness of imperforate rracheary elements, J.Lm. Provenances of specimens are given in Materials and Methods. stems) is one of the purposes of the present study. Caryophyllaceae are a family of crucial importance in demonstrating if wood anatomy can reveal whether a phylad is primarily or secondarily woody.
Woody Caryophyllaceae mostly occupy relatively dry habitats. Interesting in this connection is whether wood of Caryophyllaceae is markedly xeromorphic (for criteria of wood xeromorphy, see Carlquist 1988), and if so, in what ways. Insular representatives of a group usually tend to be less xeromorphic in various features than their continental relatives. Woody Caryophyllaceae are well represented on the Hawaiian Islands and the Canary Islands, so there is sufficient material to provide a comparison between woods of insular and continental representatives. MATERIALS  Sliding microtome sections were successfully obtained for specimens of Polycarpaea divaricata, Silene struthioloides, and all species of Schiedea. These woods are relatively uniform in texture and are moderately hard. Other caryophyllaceous woods contain hard and soft portions intermixed, or (as in the roots of Silene hawaiiensis and S. verecunda) soft tissue only. The problems posed by woods such as these were successfully solved by use of ethylene diamine for softening followed by sectioning in paraffin (Carlquist 1982). Sections were stained in a safranin-fast green combination. Both paraffin sections and sliding microtome sections were used for scanning electron microscope (SEM) study. The former were mounted on aluminum stubs just as paraffin sections are mounted on glass slides, whereas the sliding microtome sections for SEM study were dried between glass slides before being placed on aluminum mounts with the aid of double-sticky tape. Macerations were prepared with Jeffrey's fluid and stained with safranin.
Terminology follows that of the IA WA Committee on Nomenclature (1964) except for the terms vascular tracheid and vasicentric tracheid, which follow my definitions (Carlquist 1985), which I believe correspond to the intent of the IA W A Committee on Nomenclature (1964) as wel1 as to the usage of Metcalfe and Chalk (1950). The term "pervasive" axial parenchyma is proposed here for axial parenchyma present as the ground tissue of fascicular xylem instead of fibrous tissue (imperforate tracheary elements). This term was introduced in an earlier paper (Carlquist 1993). Vessel diameter is measured as lumen diameter at wide~t point.

Growth Rings
Because woody Caryophyllaceae mostly inhabit dry sites that experience only brief rainy seasons, growth rings are present in stems of most of the species studied. The only exceptions to this (habit and habitat correlations in parentheses) occur in Alsinidendron trinerve (understory shrub in wet forest); Gypsophila patrinii (root); Schiedea ligustrina ( Fig. 13) (moderated island climate); Scopulophila ri.ifordii (root); Silene hawaiiensis stem and root (moderated island climate, moderately moist zone with minimal seasonality); and S. verecunda (Fig. 25) (root). The lack of seasonality in wood of roots is likely related to succulence of roots in Caryophyllaceae, although these roots occur in some species of highly seasonal habitats.
Growth rings in Gymnocarpos salsoloides occur only occasionally (not visible in the portion shown in Fig. 1). Growth rings in this species take the form of 3 one or two layers of marginal axial parenchyma. The leaves in this species are succulent, which may be related to lack of marked fluctuation in vessel diameter. Succulents at large typically have wood in which vessel diameter does not fluctuate greatly; narrow vessels (which provide conductive safety) and wide vessels (which accommodate larger volumes of flow during periods of active growth) are of selective value in species with broader, thinner leaves. Polycarpaea divaricata, also from a dry, frost-free lowland Canarian locality, has nonsucculent leaves and moderate fluctuation in vessel diameter (Fig. 6). A similar condition was observed in Schiedea adamantis. More pronounced fluctuation in vessel diameter (as well as difference in fiber-tracheid staining capabiiity between earlywood and latewood) occurs in S. lydgatei (Fig. 17).
Minimal fluctuation in vessel diameter in Dianthus caryophyllus (Fig. 26) may be related to .the cultivated provenance of the specimen. Earlywood in D. rupicolus and Silene fruticosa consists of wider vessels in a background of libriform fibers, while latewood contains narrow vessels and a few vasicentric tracheids in a background of axial parenchyma. Vessel diameter fluctuates more markedly in D. sp. (Tobey 2026), which is a specimen from a wild area; growth rings in this species are much like those in stems of Arenaria macradenia or Silene spergulifolia. Arenaria macradenia ( Fig. 11) has very narrow vessels and vascular tracheids formed in parenchymatous background in latewood. The parenchymatous ground tissue continues in earlywood in which wide-lumened vessels ate formed. Libriform fibers supplant axial parenchyma as vessel diameters lessen within a growth ring. The same features are shown even more dramatically in the steins of Silene spergulifolia (Fig. 18). In the first two years of secondary growth in this stem ( Fig. 19), however, libriform fibers are not formed and vessels occur in a ground tissue of axial parenchyma. Vessel diameter in Silene struthioloides (Fig. 22,23) fluctuates little within a single year. Rather, vessels are narrower or wider within a particular year's growth, presumably dependent on water availability.
Presence or absence of fibers in particular stems of Caryophyllaceae has been noted by Metcalfe and Chalk (1950). In the portion of an older horizontal stem of Dianthus caryophyllus studied here, libriform fibers are present throughout most of the wood, but are absent in the most recently formed portions. The reverse was observed in stems of Schiedea globosa H. Mann (collected at Hanauma Bay, Oahu), a more nearly herbaceous plant than other species of Schiedea. In the upright flowering stems of this plant, secondary xylem begins with vessels within a parenchyma background; later-formed secondary xylem consists wholly of libriform fibers and vessels.

Vessel Elements
Vessel grouping (Table 1, column 1) is minimal (fewer than 1.20 vessels per group) in Gymnocarpos salsoloides ( Fig. 1, 2), the only species in this study in which tracheids rather than libriform fibers form the ground tissue of the wood. In· all· other Caryophyllaceae, fiber-tracheids or libriform fibers are present and vessel grouping is greater. Alsinidendron trinerve (Fig. 12) has 1.53 vessels per group, but the figure is higher in the remaining Caryophyllaceae other than Gymnocarpos. Groups of vessels are almost indefinitely large in Scopulophila rixfordii (Fig. 9; no libriform fibers are present in xylem, only narrower and wider vessels). Vessel grouping in woods of Caryophyllaceae often takes the form of radial chains (Fig. 12,13,17,23), although pore multiples are also present, especially in latewood (Fig. 11,18).
Mean vessel diameter (Table 1, column 2) ranges from a low of 8 f.Lm in Silene spergulifolia (Fig. 18,19) and 9 f.Lm in S. struthioloides (Fig. 22,23) to 26 f.Lm in Polycarpaea divaricata (Fig. 6) and Schiedea lydgatei (Fig. 17) and 27 f.Lm in Gymnocarpos salsoloides ( Fig. 1, 2). One should note that the wider vessels of Caryophyllaceae are still relatively narrow . compared to the vessels of dicotyledons at large (trees of wet forests typically have a mean vessel diameter in excess. of 100 f.Lm). The average vessel diameter figure for Caryophyllaceae studied here, 19 f.Lm, is comparable to that for Californian desert shrubs, 25.9 f.Lm (Carlquist and Hoekman 1985: vessel diameter in that study computed from outside dimension rather than lumen width). The Caryophyllaceae sampled include wood of mesic species and wood of succulent roots, and while neither category qualifies as mesomorphic, wood of Caryophyllaceae as a whole is strongly xeromorphic.
Vessel density (Table 1, column 3) has often been thought to show an inverse relationship (admittedly with exceptions of various degrees) to vessel diameter. Vessel density is notably high (above 300 per mm 2 ) in most Caryophyllaceae (Fig. I, 2 , 5, 9, 11, 13, 22, 25, 26). For those species with wide rays (e.g., Silene struthioloides) or successive cambia (Scopulophila rixfordii), rays and conjunctive tissue were not included in the areas scanned for vessel density data collection. Relatively low vessel density (55 per mm 2 ) was observed in the succulent roots of Silene hawaiiensis. In all other species, vessel density was within a range that is typical of more xeromorphic woods. ALISO Vessel element length (Table I, column 4) is notably short in Caryophyllaceae, as shown for Gymnocarpos salsoloides in Fig. 3. The longest mean vessel element length (218 f.Lm) was observed in the species from the most mesic locality, Alsinidendron ttinerve. Mean vessel element length in this species is not markedly greater than that for Caryophyllaceae as a whole (150 J.Lm), however. In tum, the figure for the family as a whole is not much greater than that in the species with shortest vessel elements: Silene spergulifolia (113 f.Lm) and Dianthus caryophyllus (114 f.Lm).
The vessel-to-vessel and vessel-to-imperforate tracheary element pits are ·basically alternate in Caryophyllaceae and typically range from 4 to 5 f.Lm in vertical dimension of the pit cavity (Fig. 4, 7); larger pits (6 J.Lm) were observed only in Alsinidendron trinerve. Later(\lly elongate pits characterize vessels of Arenaria macradenia, Dianthus caryophyllus , Gypsophila patrinii (Fig. 30), Scopulophila rixfordii, Silene spergulifolia (Fig. 21), S. struthioloides, and S. verecunda. These laterally elongate pits are termed pseudoscalariform here: in shape they are like those of scalariform pitting in being laterally elongate, but rather than lining up in vertical series on wall faces, they are staggered, illustrating that they may have been derived from lateral elongation of alternate pits. Pseudoscalariform pits are not subdivided by narrow strands of secondary wall material in Gypsophila patrinii (Fig. 30); there are slender strands that run vertically across a few pits in vessels of Dianthus caryophyllus (Fig. 29), and wider strands across a few pits in vessels of Silene spergulifolia (Fig. 21). The list of species with pseudoscalariform pitting is the same as the list of species with pervasive vessel-to-parenchyma pitting. In Dianthus rupicolus and Silene fruticosa, pits on the latewood vessels are somewhat larger, and with bigger apertures, than pits on the earlywood vessels (which face mostly libriform fibers rather than parenchyma, but pitting is alternative, not really pseudoscalariform. Pseudoscalariform pitting occurs in vessel- to-parenchyma pitting; in some of the species with these pits, libriform fibers occur in some zones, yet pseudoscalariform pitting on vessels occurs throughout the wood. In Silene struthioloides, which has pseudoscalariform pitting, vessels are embedded in a background tissue of libriform fibers (Fig. 24 ). The woodiness of this species is exceptional in the genus, and the pitting pattern may have been retained from an ancestry in which parenchyma was abundant in fascicular secondary xylem.
Helical sculpture is not evident in vessels of Caryophyllaceae, with a few notable exceptions. Pairs of faint thickening bands accompany pit apertures in vessels of stems of Silene hawaiiensis. These are most clearly seen in narrow vessels. With light microscopy, helical sculpturing in vessels of Gymnocarpos salsoloides is so faint it would most likely be regarded as dubiously present by most observers. With SEM, thickenings are clearly present, although they are not pronounced (Fig. 5). Helical sculpture was also observed with SEM in vessels of Schiedea ligustrina (Fig. 15, 16). Curiously, in the example of more pronounced thickenings shown in Fig. 16, the thickenings run counter to the direction of the pit apertures.

Imperforate Tracheary Elements
In Caryophyllaceae as a whole, length of imperfomte tracheary elements (Table l, column 6) averages 257 JJ.m, which is 1.71 times the average vessel element length (150 JJ.m) for the species studied. Woods of some species lack imperforate tracheary elements: roots of Gypsophila patrinii, Scopulophila rixfordii, and Silene verecunda.
Caryophyllaceae are striking in terms of diversity of 9 imperforate tracheary elements. Gymnocarpos salsoloides has fully bordered pits on imperforate tracheary elements (Fig. 3, 4). These cells are therefore tracheids according to the lA WA Committee on Nomenclature (1964 ). The mean pit cavity diameter of these tracheids is 4.2 1-Lm. Tracheids like these have been reported in Gymnocarpos decandrum Forssk. by Fahn et al. (1986). Fiber-tracheids that have pits with vestigial borders 0.5-1.5 JJ.m in diameter were observed in Alsinidendron trinerve, Polycarpaea divaricata, and all of the species of Schiedea. The species studied in Arenaria, Dianthus, and Silene lack borders on pits of imperforate tracheary elements, which are therefore termed libriform fibers.
A few layers of vascular tracheids were observed in the latewood of Schiedea lydgatei. These are inconspicuous and easily overlooked. In Silene spergulifolia, the cells that appear to be extremely narrow vessels in latewood (e.g., Fig. 18, top) when examined in longitudinal sections and in macerations usually lack perforation plates and thus are mostly vascular tracheids. On the other hand, the wood of Scopulophila rixfordii (Fig. 9) has what appear to be very narrow vessels randomly intermixed with wider ones. Some of these apparently narrower vessel elements do have perforation plates, but others do not. These latter cells should be termed vasicentric tracheids because of their distribution.

Axial Parenchyma
Pervasive axial parenchyma occurs in the roots of Gypsophila patrinii, Scopulophila rixfordii (Fig. 9), and Silene verecunda (Fig. 25). As noted by Metcalfe  (Fig. 13, 14, scale above Fig. 1; Fig. 15 and Chalk (1950), certain other Caryophyllaceae have axial parenchyma in bands that alternate with the parenchyma bands. These parenchyma bands can be called marginal parenchyma in some species because they are positioned at the end and beginning of growth rings, as in Arenaria macradenia (Fig. 11 ), D. rupicolus, Dianthus sp. (Tobey 2026), Silenefruticosa, and S. spergulifolia (Fig. 18). The parenchyma bands in Arenaria macradenia, D. sp., and S. spergulifolia include the narrow latewood vessels as well as large earlywood vessels; midseason vessels of intermediate diameter are embedded in a ground tissue of libriform fibers. Axial parenchyma formed in a less predictable fashion, with no evident relationship to growth rings, was observed in Dianthus caryophyllus (Fig. 26). Marginal parenchyma consisting of bands only a single cell wide (rarely two) was observed in Gymnocarpos salsoloides, and was reported in G. decandrum by Fahn et al. (1986). In addition to banded axial parenchyma, scanty vasicentric parenchyma characterizes many Caryophyllaceae; it is seen clearly in Gymnocarpos salsoloides (Fig. 2). Occasional diffuse cells can also be seen in that species (Fig. 2). Scanty vasicentric parenchyma was recorded for Alsinidendron trinerve (in which the cells have thin primary walls only), Polycarpaea divaricata, and all species of Schiedea (very rare in S. ligustrina and S. lydgatei), and Silene struthioloides. In the earlywood of Dianthus rupicolus, in which Iibriform fibers occur, no axial parenchyma was observed adjacent to vessels. In Silene fruticosa, some of the earlywood vessels embedded in the libriform fibers are associated with a few vasicentric axial parenchyma cells.
Axial parenchyma of Caryophyllaceae as seen in longisection usually consists of cells that are not subdivided into strands. Strands of 2-3 cells occur in Gymnocarpos salsoloides and were reported for G. decandrum by Fahn et al. (1986). Starch was observed in axial parenchyma of Schiedea lydgatei.

Rays
The generalization by Metcalfe and Chalk (1950) that woods of Caryophyllaceae are rayless does not hold for all species. It does hold, in the present study, for Alsinidendron trinerve (Fig. 12), Schiedea (all species: Fig. 13, 14, 17); and stems of Silene hawaiiensis. In the wood of Dianthus rupicolus, wood begins rayless, but in the outer wood, rays do finally originate. ALISO A few uniseriate rays one or two cells high were observed in the wood of Silene fruticosa, which is otherwise rayless.
In Arenaria macradenia and Gypsophila patrinii, multiseriate rays are relatively common (mean height in A. macradenia, 529 j.Lm). Multiseriate rays are also relatively common in Silene struthioloides (Fig. 22). Silene struthioloides has sufficient accumulation of secondary xylem so that one can see origins of rays. The multiseriate rays in this species and in Dianthus rupicolus originate abruptly, rather than by gradually widening of uniseriate or biseriate rays. Abrupt origin of multiseriate rays has been observed in Cucurbitaceae (Carlquist 1992) and Aristolochiaceae (Carlquist 1993). Wide multiseriate rays were also observed in roots of Caryophyllaceae in Silene hawaiiensis and S. verecunda. In all of the multiseriate rays as studied in radial section, ray cells are upright and thin walled, but a few procumbent cells were also seen in rays of Arenaria macradenia and Silene struthioloides.
In Dianthus arboreus L., Tellini (1939)  In Gymnocarpos salsoloides, there are inconspicuous uniseriate (rarely biseriate) rays (Fig. 3 ). These . Dianthus caryophyllus (Carlquist s.n.), sections of older horizontal stem.-26. Transection showing secondary xylem below from normal cambium above; below in photograph cambium has formed in the pith, yielding xylem (two dark patches) to the outside, phloem to the outside, and a periderm (bottom of photograph).-27. Tangential section through a zone of axial parenchyma, showing rayless condition.-28. Portion of tangential section showing vessels with laterally elongate pits.-29. SEM photograph of vessel from tangential section showing laterally elongate (pseudoscalariform) pits traversed by a few narrow strands of wall materiaL (Fig. 26, scale above Fig. I; Fig. 27, seale above Fig. 2; Fig. 28, scale above Fig. 4; Fig. 29, bar at upper right 5 j.Lm.) average 189 11-m in height and are rarely more than three cells in height. Cells of these rays are upright, and their tips often overlap as seen in tangential section. Uniseriate rays of this description were reported in Gymnocarpos decandrum by Fahn et al. (1986). A few uniseriate rays like those of Gymnocarpos salsoloides were observed in the wood of Polycarpaea divaricata.

Crystals
Druses are present in woods of the family (Fig. I 0,  31). The druses of Caryophyllaceae, as studied by SEM, are distinctive in that they are composed not of rhombohedral (prismatic) crystals but of chip like or wedge like fragments (Fig. I 0, 31 ).

Storying
Patches of storied axial parenchyma and vessel elements, sometimes with storied libriform fibers as well, were observed in tangential sections of wood of Dianthus caryophyllus, Polycarpaea divaricata, and Silene spergulifolia. Storying has not hitherto been reported in Caryophyllaceae, but it has been reported in three other families of Caryophyllales: Amaranthaceae, Chenopodiaceae, and Nyctaginaceae (Metcalfe and Chalk 1950). The species of Caryophyllaceae that exhibit storied wood structure are precisely the ones that have the shortest vessel elements. Vessel element length is extremely short in species with storied structure in dicotyledons at large (Bailey and Tupper 1918;Bailey 1923). Pfeiffer (1926) reports cambial variants for the family; I presume these to be successive cambia forming strands of vascular tissue like those of Scopulophila rixjordii (Fig. 9). The tribe Paronychieae is richest in this cambial type, whereas it has been reported in only one species of subfamily Silenoideae (Pfeiffer 1926). All instances of successive cambia reported thus far are in roots, except for three genera which have successive cambia in both stems and roots: Cometes, Corrigiola, and Pollichia (Pfeiffer 1926). ALISO A unique form of successive cambial activity is newly reported here for horizontal older stems of Dianthus caryophyllus (Fig. 26). At the periphery of the pith in this material, a cambium produces vascular tissue in inverse orientation-phloem to the inside and xylem to the outside. In addition to producing phloem on its internal face, this cambium yields a periderm internal to the phloem (Fig. 26, bottom).

ECOLOGICAL CORRELATIONS
Vessel features demarcate Caryophyllaceae as markedly xeromorphic: vessel elements are short, narrow, and numerous per mm 2 of transection (for further comments on wood xeromorphy, see Carlquist 1988). The features for all Caryophyllaceae studied here averaged are as follows (in parentheses, figures for Californian desert shrubs from Carlquist and Hoekman 1985): vessels per group, 3.14 (2.62); vessel diameter, 19 11-m (26 11-m); vessels per mm 2 , 372 {361); vessel element length~ 146 11-m (190 1-1-m).In all four features, the Caryophyllaceae sample studied here yields means indicative of more xeromorphy than the means for the Californian desert shrubs, despite the fact that only a few of the species studied come from localities as dry as desert sites, namely Arenaria macradenia, Gymnocarpos salsoloides, and Scopulophila rbfordii. One species studied here, Alsinidendron trinerve, comes from a frequently cloudy wet forest area where it is an understory shrub. Although the wood of Alsinidendron is less xeromorphic than that of the closely related Schiedea in quantitative features of vessels, the difference is not great despite the fact that the Schiedea species were collected in dry lowland localities.
Another comparison of potential interest is between insular and mainland Caryophyllaceae. This comparison is feasible because so many insular species of Caryophyllaceae have been utilized in the present study. Dianthus rupicolus and Silene fruticosa have been omitted from this comparison because their ecology is maritime continental-intermediate between insular and inland continental-and because the samples were from cultivated specimens, which one would expect to have more xeromorphic wood expressions than wildgrowing plants. Quantitative vessel features of insular species averaged are as follows (means for combined continental species in parentheses): number of vessels per group, 1.76 (2.34); vessel diameter, 20.6 11-m (14.0 11-m); number of vessels per mm 2 , 304 (605); vessel element length, 153 11-m (143 11-m). The insular species are more mesomorphic than the continental species in all of these features. Another way to express this is to combine three vessel features into the Mesomorphy ratio (vessel diameter times vessel element length divided by number of vessels per mm 2 ). The figure for the insular species for this ratio is 10.37, whereas the continental species yield the value 3.31. While one can say that the woods of insular species show less xeromorphy, as might be expected if insular climates are less extreme, the difference is not huge. One should note that some insular species grow in localities that are in effect deserts: Gymnocarpos salsoloides (Teno, Tenerife) and Silene struthioloides (Haleakala Caldera, Maui).
Other putatively xeromorphic features that occur in the wood of Caryophyllaceae include helical sculpturing in vessels and type of imperforate tracheary element. Helical sculpture occurs in vessels of Gymnocarpos salsoloides, Polycarpaea divaricata, and Schiedea ligustrina. These species do not have highly pronounced helical sculpturing in vessels. Nevertheless, these occurrences are of great interest, because helical sculpturing is sometimes thought to be associated only with cold, and can be found in some genera (e.g., Cornus) that never occur in dry areas. In fact, helical sculpturing may occur in species that grow in cold areas in some instances, but in other groups the habitat is a dry one. The three species of Caryophyllaceae with helical sculpturing in vessels occur in areas that experience no frost.
Gymnocarpos has tracheids rather than fiber-tracheids or libriform fibers. It also has a low number of vessels per group ( 1.1 0). This is one more instance that affirms that vessel grouping is minimal (essentially restricted to random contacts between vessels governed by vessel density and vessel diameter) if tracheids are the type of imperforate tracheary element present. Tracheids can serve as a conductive system subsidiary to vessels, theoretically serving for conduction should the vessels embolize (Carlquist 1984); the conductive capabilities of tracheids are sufficient to render the selective value of vessel grouping minimal. According to this hypothesis, fiber-tracheids and libriform fibers are nonconductive or essentially so, so that vessel grouping offers a redundancy (while maintaining conductive pathways) valuable should some vessels embolize. Vessel groupings should be greater the more xeromorphic the locality where such woods are involved. All Carophyllaceae other than Gymnocarpus have fiber-tracheids or libriform fibers, and of these species, the least degree of vessel grouping does occur in the species from the most mesic locality, Alsinidendron trinerve. A low degree of vessel grouping in Di-anthus caryophyllus may indicate the greater degree of water availability in cultivation.
Growth rings in Caryophyllaceae do not show dramatic differences in vessel diameter in the range of what one sees in temperate species of Quercus. However, most localities inhabited by woody Caryophyllaceae have a wet season with short availability of a modest quantity of moisture. The most pronounced growth rings are in two species of desert areas, Arenaria macradenia and Silene spergulifolia. Roots tend to store water in those Caryophyllaceae with perennial woody roots, and this should lessen the likelihood of growth ring formation. Indeed, growth rings are absent in the roots of Gypsophila patrinii, Scopulophila rixfordii, Silene hawaiiensis, and S. verecunda.

HABITAL CORRELATIONS
Caryophyllaceae are best described, when maximally woody, as subshrubs; none qualifies as a shrub in the sense of shrubby Fabaceae or Rosaceae. Woody Caryophyllaceae could, in fact, be regarded merely as woody herbs. The occurrence of raylessness is pertinent in this regard. Raylessness has often been associated with phylesis from herbaceousness toward increased woodiness (Barghoorn 1941;Carlquist 1970Carlquist , 1988. In Caryophyllaceae, raylessness is most pronounced in Alsinidendron and Schiedea (the latter perhaps the woodiest genus of the family). Uniseriate rays without multiseriate rays characterize Gymnocarpos and may be found sparsely in Polycarpaea. Multiseriate rays appear characteristic of the majority of remaining Caryophyllaceae studied here, except for the stems of Dianthus caryophyllus (a few multiseriate rays) and Silene spergulifolia (no rays). One can conclude that Caryophyllaceae are by no means a rayless family, and that if raylessness is an indicator of secondary woodiness, the family might be primitively woody, but with evolution of herbaceousness in most of the family and secondary woodiness in some phylads. Successive cambia, reported here in Scopulophila and Dianthus, and reported in 19 genera of the family by Pfeiffer (1926), have been regarded by some workers as an indication of secondary woodiness in phylads that mostly lack normal cambial activity (see discussion in Carlquist 1988). This is probably not true in typically woody phylads (e.g., particular species of Bauhinia of the Fabaceae are characterized by either a normal cambium or successive cambia). The raylessness, lack of either multiseriate rays or uniseriate rays, abrupt origin of multiseriate rays, and occurrence of successive cambia in Caryophyllaceae may be indicative of loss of particular features of normal cambial activity. Normal cambial activity is not a single phenomenon, but a series of phenomena. Loss of capability for production of rays, or production of par-ALISO ticular kinds of rays, can occur without loss of capability for production of axial tissue, for example.
The occurrence of libriform fibers in Arenaria, Dianthus, and Silene spergulifolia is noteworthy in that it seems related to habit. Libriform fibers in the horizontal basal stem of Dianthus caryophyllus may cease to be formed once maximal stem diameter has been attained, if the stem studied is typical: optimal mechanical strength has been attained by that time, but production of more conductive tissue is required. In Schiedea globosa, the reverse sequence is characteristic (vessels in a parenchyma ground tissue, followed by fiber-tracheids without parenchyma or vessels) of that species, and may correlate with the fact that the stem, as it produces an inflorescence and fruits, has a progressively increased requirement for mechanical strength.
Successive cambia in the roots of Carophyllaceae may relate to the probable water and photosynthate storage functions of these roots. Successive cambia are a way of producing vascular tissue scattered throughout a parenchymatous ground tissue (the conjunctive tissue). If so, this would account for the fact that successive cambia are much less common in stems than in roots of Caryophyllaceae. Storage function could be achieved in alternative ways, such as bands of libriform fibers alternating with bands of (marginal) parenchyma (Arenaria, Silene spergulifolia) or pervasive parenchyma (roots of Gypsophila patrinii and Silene verecunda).

CORRELATIONS
The interrelationships of families of Caryophyllales have interested numerous workers in recent years. The cladistic work of Rodman et al. (1984) places Caryophyllaceae near the base of the caryophyllalean clade. Bedell ( 1980) claimed a relationship between Stegnosperma (Stegnospermaceae) and Caryophyllaceae. The recent assemblage of rbcL data (Chase et al. 1993) shows Caryophyllaceae as a branch from well along the Caryophyllales clade, near Chenopodiaceae and Amaranthaceae.
Further knowledge of caryophyllalean woods is desirable before defensible conclusions can be reached. However, the present study does demonstrate the diversity of woods within the family, a key point in establishing its relationship. The wood of Gymnocarpos is of significance in this regard, because its tracheids and uniseriate rays are unlike the fiber-tracheids (Alsinidendron, Polycarpaea, Schiedea) or libriform fibers (other genera) and multiseriate rays (or raylessness) elsewhere in the family. The features of Gymnocarpos wood support the idea that Caryophyllaceae branched early from the caryophyllalean clade, other-wise one would not expect tracheids or even fibertracheids in the family, and one might not expect rays. Wood of Amaranthaceae and Chenopodiaceae (libriform fibers, raylessness, storying, and with successive cambia rather than a single cambium), and the range of character-state expressions in caryophyllaceous wood could not be derived from that pair of families unless one hypothesized a large number of reversions to more primitive expressions. Wood features tend to be less reversible than other characters because of functional correlations between the wood characters (Carlquist 1988). Stegnosperma does have tracheids, but it also has successive cambia and both multiseriate and uniseriate rays. For Stegnospermaceae and Caryophyllaceae to be related, one would have to imagine an ancestor with tracheids, but with multiseriate and uniseriate rays (instead of only uniseriate rays, as in Gymnocarpos) and yet with no successive cambia. Genera with such ancestral wood features do exist in other families of Caryophyllales. At present the cladistic scheme of Rodman et al. (1984) is supported by the wood data.
The ancestors of Caryophyllaceae are likely to have been woody and to have had rays. If the entirety of Caryophyllaceae were rayless or had successive cambia, one could hypothesize ancestors to Caryophyllaceae that had lost normal cambial activity. The woods of Gymnocarpos and Polycarpaea suggest a normal woody ancestry for the family, even if they are not big shrubs. Interestingly, the Pax and Hoffmann (1934) system placed Gymnocarpos and Polycarpaea near the beginning of Paronychieae, the tribe they recognize as most primitive, although they had no wood data. The presence of tracheids rather than fiber-tracheids in Gymnocarpos also is in line with a basal placement in the family. All of the instances of successive cambia known thus far are in Paronychia, with the exception of a single species of Silene (Pax and Hoffmann 1934). The distribution of this feature in Caryophyllaceae needs more study, particularly with respect to roots.
Fiber-tracheids with much reduced pit borders are found in Alsinidendron and Schiedea, which fall into the central tribe of Caryophyllaceae, Alsineae, in the Pax and Hoffmann (1934) system. The raylessness of these two genera distances them from Gymnocarpos and Polycarpaea. However, Arenaria, which has multiseriate rays and nucleate libriform fibers, is also placed in Alsineae.
The third, and presumably most specialized, tribe in the Pax and Hoffmann system is Sileneae, and all of the genera of that subfamily studied here have libriform fibers rather than tracheids or fiber-tracheids. Wide multiseriate rays that originate abruptly occur in Sileneae but also in one genus of Alsineae (Arenaria).