Wood and Bark Anatomy of Myricaceae: Relationships, Generic Definitions, and Ecological Interpretations

Wood anatomy of the single spec ies of Cana comyrica (hitherto not studied) shows that it belongs in Myricaceae, although it differs from other genera in several respects (axial parenchyma grouped in bands or columns as well as diffuse ; Heterogeneou s Type I rays; more numerous bars per perforat ion plate). The latter two features are primitive for the family . The four genera (Canacomyrica. Comptonia, Morella, and Myrica s.s .) differ from each other not onl y by qualitative features but by qu antitative features (feature means in genera mostly nonoverl apping). Wood of Comptonia and Myri ca s.s. lacks chambered crystals in axial paren chyma and ray cr ystals. Wood of Myrica S.s. has tracheids in latewood but fiber-tracheids in earl ywood. Diagnostic generic summaries are presented. Features of Myricaceae such as scalarifonn perforation plates, presence of (true) tracheids , ray types , chambered encapsulated crystals in axial parenchyma, and bark anatomy correspond with character states and expressions in Betulaceae, Casuarin aceae, Corylaceae, Juglandaceae (including Rhoipteleaceae), Ticodendrace ae and, to a lesser extent , Fagaceae and Nothofagaceae. This grouping of families can be found as Fagales in recent DNA trees . The predominance of tracheids in basal Fagales such as Myricaceae and Ticodendraceae suggests that origin of vasicentric tracheids which occur in combination with libriform fibers in Fagaceae is the product of tracheid dimorphism. Low imperforate tracheid length to vessel element length rat ios (FN ratio s) in Myricaceae are a probable indication of wood primitiveness. Quantitative vessel features of Myricace ae, as combined in Mesomorphy Ratio values , characterize wood of Myricaceae as a whole, but at the species level such values correspond to respective habitats; notably high vessel density in Comptonia may represent greater conductive safety appropriate to relatively dry habitats .


INTRODUCTION
No monographic study of wood anatomy of Myricaceae exists, although details of wood of a scattering of species of Myricaceae can be found in a number of publications (see Gregory 1994) .A condensed summary based on a small number of species can be found in Metcalfe and Chalk (1950).The present study is based on specimens of all species available in leading xylaria, plus collections by the writer.This monograph is intended to address particular questions about the systematics of Myricaceae and Fagales, but also to analyze the data in terms of ecology, habit, and particular anatomical modes of structure.
Definitions of genera in Myricaceae have varied considerably.The inclusion of Canacomyrica has been questioned (Thorne 1968(Thorne , 1976)), although no alternative placement has been proposed (Cronquist 1981 ; Thorne 200 I).Wood anatomy of Canacomyrica has not been described hitherto.Comptonia seems universally recogn ized as a monotypic genus (Elias 1971 ;Cronquist 1981; Thorne 200 I) .Likewise, a number of workers have treated the species sometimes formerly known as Gale hartwegii A. Chev.and G. palustris A.
Chev. but more commonly under Myrica as a genus separate from the remainder of the family .The correct name for this small genus is not Gale, as assumed by Hylander (1945), who mi stakenly thought that Myrica cerifera L. is the type species of Myrica (see Elias, 1971, p. 309), but Myrica, because the type of the genus is actually Myrica gale L. Elias (1971) recognizes sections Gale and Morella as sections of Myrica, but more recent workers have raised these sections to genera.Morella is being recognized by taxonomic workers (Knapp 2002), and in floristic treatments such as Killick et al. (1998) and Goldblatt and Manning (2000).Although in many families , wood features may follow ecological adaptations primarily, and relatively few criteria are present in wood features for generic distinctions, that proves not to be true in Myricaceae.The present study thus becomes an excellent example of sy stematic wood anatomy, in which all of the genera are amply distinguishable in terms of wood feature s.Unfortunately, all of the recognized genera have not been samp led for the purpose of constructing DNA-based phylogenetic trees, and such data is much needed.DNA data is available for only a tiny fraction of the estimated 45 valid species of Myricaceae (Thorne 200 I ).
Myricaceae have been generally placed in Fag ales or Juglandales, if recognized separately from Fagales (summary of treatment of 12 phylogeneticists in Goldberg 1986).Molecular dat a have confirmed this pla cement.Analyses offer tree s that place Myricaceae in the rosid line in a clade that corresponds to Fagales; closeness to Jugland aceae is indi cated (e.g., Soltis et al. 2000), but some degree of proximity to Betulaceae and Casu arinaceae is also possible (Manos and Steele 1997).If floral morphology is a reliable indicator of relationships, fem ale flowers of Jugland aceae are worthy of considerati on.Just as the female flowers of Jugland aceae ha ve inferior ovaries (C ronquist 1981), those of Myri caceae also appear to be epigynous (MacDonald 1989).In Canacomyrica (Fig. I), this seems to be true, because a crown of five teeth, presuma bly calyx teeth, crown the fruit.Alternatively , the se have been interpreted as a late-developing dis c that envelops the ovary (Cronquist 1981).Three branched stigmas (dark in Fig .I) tip the fruit of Canacomyrica.Between the cal yx teeth and the stigmas are stame ns, rendered visible in Fig .I by their covering of pale golden glandular trichomes (some of these trichomes are also seen on the bract below the ovary).MacDonald (1989) figures apparent api cal bract s, like the putati ve calyx teeth of Canacomyrica, for Myrica gale.The api cal androecium of Canacomy rica has a counterpar t in the apical androecium figured for Morella califo rnica (MacDonald 1989 ). Th e fem ale flowers of Canacomyrica show at least supe rficial similarity to flow er s of pomoid Rosaceae, if the above intepretation is correct.MacDonald (1989) interprets Comptonia as unique among Myricaceae in having a superior ovary.Developmental studies of myricaceous flowers are very mu ch needed.
The notion that Canacomyrica do es not have orthotropous ovules was corrected by Leroy (1949Leroy ( , 1957)), who also demonstrated that wh at Guillaumin, in describing the ge nus, thought was an elongate funiculus is actually a prolonged integument (see Cronqui st 1981, for a full account ).Sundberg (1985) has concluded that pollen of Canacomyrica is compatible with that of other Myricace ae, althou gh generica lly distinct.
Thus, comparisons of wo od of Myricaceae with that of oth er Fagales, es pecially Ju glandaceae, are appro-priate.The submers ion of Rho ipteleaceae into Jugl anda ceae (e.g ., Thorne 200 I) suggests that Rhoiptelea must be included in comparisons also.Be cause the fagalean clade is a segment of the larger gro up now generally termed eurosid s I (e.g., Soltis et al. 2000), Rosace ae might be expected to share so me features with Fagales as a whole and Myricaceae in particular.The controversial matter of agg regate rays and compound rays , whi ch are common in some Fagale s (Metcalfe and Chalk 1983 ) mu st ine vita bly be considered in terms of whether Myricaceae offer any equivalents to such rays.Likewise, the relationship between tracheids in Myricaceae and vasicentric trach eid s in Fagaceae needs examination.
Myricaceae occupy a range of habitats, appare ntly mostly mesic.Youngken (191 9) claims sw ampy habitats for all eastern U.S. species except for Comptonia, which " thrives in dry sterile soi l." Morella californica occ urs in canyons and moist slopes, wh ere as Myrica hartwegii is found on montane stream banks (Munz 1959).The South African spe cies (Morella ) occur on sandstone or on sa ndy or limestone slopes (Goldblatt and Mann ing 2000), but thos e authors do not describe water ava ilability in these localities.Canacomyrica grows in moist for est near the summit of the Montagne des Sources, New Caledonia (orig inal observation ).The range in quant itat ive vesse l data within the family is conside rable, ind icating that adaptation to a range of habitats has occurred, so an alyses like those of Carlqu ist and Hoekman (1985 ) are undertaken.M. hartwegii, also the genus GaLe of some authors) is termed Myrica S.s.(sensu stricto) here so as to distinguish it from the more inclusive Myrica of many earlier authors.

Sp
The wood and bark samples were all available in dried form.Samples were boiled and stored in 50% aqueous ethanol.Woods of Myricaceae are not exceptionally hard or soft, so they can be sectioned on a sliding microtome without pretreatment to mitigate hardness.Bark adhered to some wood sections and provided usable data on bark anatomy.Sections for permanent slides were stained with safranin and counterstained with fast green.Some sections were left unstained , dried between clean glass slides, mounted on aluminum stubs, sputter-coated, and examined with a Hitachi 2600N scanning electron microscope (SEM).Macerations were prepared with Jeffrey 's Fluid and stained with safranin.
Mean number of vessels per group is based on an average of counts in which a solitary vessel = I, a pair of vessels in contact = 2, etc. (this can also be defined as number of vessels divided by number of vessel groups, solitary vessels in that case counting as a group).Because vessel tips overlap considerably, a transection of the overlapping region can appear as a pair of vessels, and care was taken to distinguish these instances from genuine pairs of vessels in contact.Vessel lumen diameter is measured rather than outside vessel diameter because the lumen diameter is significant hydrologically.Terms used are in accordance with the IAWA Committee on Nomenclature (1964), which distinguishes among tracheids, fiber-tracheids, and libriform fibers in dicotyledonous woods, but the term vasicentric tracheid is used in accordance with Carlquist (2001), a usage that follows closely the usage of Metcalfe and Chalk (1950), and the term "apotracheal columns" is proposed for a grouping of axial parenchyma cells found in Canacomyrica.The arrangement of photographs (Fig. 1-45) is alphabetical by genus.

Growth Rings
Most Myricaceae show growth rings, ranging from almost imperceptible, as in Canacomyrica (Fig. 4, latewood at bottom) to more pronounced, as in Comptonia (Fig. 8), Morella caLifomica (Fig. 10), M. inodora (Fig. 14), M .nagi (Fig .16), M. rubra (Fig. 38, 40), M. quercifolia, Myrica gale (Fig. 42), and M. hartwegii (Fig. 44).No appreciable growth ring formation was observed in M. javanica (Fig. 26), M. kraussiana (Fig. 30), or M. pubescens (Fig. 32).Growth rings, where minimal, feature a small change in vessel diameters between earlywood and latewood (Fig. 4,12).In species of Morella with more pronounced growth rings, the latewood is brief, i.e. , most of the growth ring contains vessels of moderate diameter and narrower vessels are formed only a short distance prior to the ends of the growth rings.In species with more pronounced growth rings, the latewood tracheids tend to be thicker walled and radially narrower than those in earlywood (Fig. 14,16,40).Myrica S.s. is distinctive in having numerous vessels in the earlywood (Fig. 42, 44).These vessels are densely placed in earlywood, and therefore, the number of vessels per group is higher in Myrica s.s.than in other Myricaceae (Table I, column 1).However, in the latewood, vessels are sparse (Fig. 44, center) or absent (Fig. 42,center).Thus, Myrica s.s.represents growth ring Type 5A or 5B (Carlquist 200 I), whereas the remaining Myricaceae fall into Type ID (vessels wider in earlywood, tracheids wider in earlywood) or Type IC (vessels wider in earlywood, but tracheids about the same throughout the ring).

Quantitative VesseL Features
Vessel grouping (Table I, column I) is minimal in Myricaceae.One must view reports of vessel grouping in dicotyledons with tracheids as the imperforate tracheary element type with care, because mistaking a transection of overlapping vessel ends as a pair of vessels is easy to do.Vessel grouping is as low as one finds in any family of vessel-bearing dicotyledons if one deducts the two species of Myrica s.s.
If one measures mean diameter of vessels as seen in transection (Table I, column 2), narrower vessels characterize Comptonia peregrina.Myrica gale, and     M. hartwegii.Narrow vessels were also observed in Morella qu ercifolia, although the small diameter (6 ern) of that wood sample and the fact that this species is shrubby rather than arboreal may account for this .The widest vessels observed in the study were those of M. javanica (Fig. 26).The transections of wood presented at the same scale (Fig. 5,8,10,12,14,16,26,30,32 ,38,42 ,and 44) show that there is an appreciable range in vessel diameters within the fam ily, and these photographs accurately reflect the mean lumen diameters recorded in Table I, column 2. Mean vessel density (Table 1, column 3) shows an amazing range from I I per mrn-(Morella javanica, Fig .26) to 260 per mm -(Comptonia peregrina, Fig. 8) .This range is greater than one might expect from the span of vessel lumen diameters in the family.This circumstance will be examined below in terms of possible physiological significance .
Vessel element length (Table I, column 4) ranges from relatively short in the s hru bs Comptonia peregrina (477 urn) and Morella qu ercifolia (462 urn) to long in trees such as M. javanica (800 urn), M. dom-   ingana (80 I J.Lm) and M. rubra (812 J.Lm).The mean for the family as a whole (641 J.Lm) is remarkably close to the mean given by Metcalfe and Chalk (1950) for dicotyledons at large, 649 J.Lm.
Vessel wall thickness (Table I, column 5) is much less (family mean, 1.8 J.Lm) than the thickness of tracheid walls, and varies relatively little within the family.The thinness of vessel walls is perceptible in the higher-scale photographs (Fig. 6, 29).
Pits on lateral walls of vessels are oval to circular and do not deviate much, throughout the family, from a diameter of 4 J.Lm (pit cavity diameter measured parallel to long axis of vessel element), as seen in Fig. 22-24.Pits are slightly larger, averaging about 5 J.Lm in diameter, in Canacomyrica monticola (Fig. 6), Morella californica, M. javanica (Fig. 28), M. pubescens (Fig. 35), and M. salicifolia.Species with pits approaching 3 J.Lm in diameter include Morella cerifera, M. faya, M. kraussiana, M. quercifolia, Myrica gale, and M. hartwegii.
Perforation plates of Myricaceae are generally sealariform, with a range of 1-32 bars (Table I, column 6).One genus, Comptonia, has simple perforation plates almost exclusively, with a very small proportion of the plates bearing one to three bars.At the opposite extreme is Canacomyrica, in which the material studied averaged 24.8 bars per plate.Second to Canacomyrica is Myrica s.s., the two species of which together average 12.6 bars per plate.All of the species of Morella fall below Myrica in average bar number (Table I,column 6;(34)(35)(36).If one averages all species of Morella studied, one obtains the figure of 8.0 bars per plate.Thus, the four genera have notably different modes with respect to this feature.
Perforation plates with fewer bars per plate sometimes have wider bars (Fig. 24) than those with more numerous bars (Fig. 22, 23), but exceptions are easy to find.Notably wide but relatively numerous bars (Fig. 34) and notably slender bars in plates with few bars (Fig. 35, 36) may readily be found.Where wider, the bars have easily seen borders (Fig. 18, 34) , and probably no bars lack borders entirely, at least if one looks at the lateral ends of the perforations.

Qualitative Vessel Features
The limits between perforation plates and lateral wall pitting are not as clear as one might think if one studies the transitional pits with SEM.Transitional pits may bear porose pit membranes indicating only partial lysis of the pit membrane (Fig. 18).Pit membranes of pits on lateral walls of vessels do not bear these pores.
Pits on lateral walls of vessels are transitional, opposite, or alternate on vessel to vessel interfaces, and types vary greatly within species.Vessel to vessel interfaces occur mostly in the overlap areas between vessels, because degree of vessel grouping is so low.Vessel to ray pitting is scalariform (infrequently), transitional, opposite, or alternate, but rather frequently opposite.On vessel surfaces in contact with tracheids, pits are opposite or alternate (Fig. 35), quite sparse in some areas (Fig. 36) .Opposite or alternate pits have pit cavities oval in face view, but the pit apertures facing the vessel lumen are narrowly elliptical (Fig. 35).
No helical thickenings or vesturing of any kind were observed in vessels of Myricaceae.
Thin-walled tyloses, often containing resinlike deposits, were observed in several Myricaceae; they were exceptionally common in material of Canacomyrica and Myrica hartwegii.

Imperforate Tracheary Elements
With the exception of the two species of Myrica.all imperforate tracheary elements in Myricaceae are tracheids.This designation is easy to make because the pit cavities are large ("fully bordered pits") and the pits are densely placed (Fig. 23, 24, 34, and especially Fig .25).Pit cavities filled with dark-colored compounds show the bordered nature of the pits in tracheids (Fig. 6, 28).The apertures facing the lumina of tracheids are slitlike (Fig. 37).
In the two species of Myrica, a curious and subtle difference between earlywood and latewood occurs with respect to imperforate tracheary elements.In addition to vessels, the earlywood contains fiber-tracheids with sparsely distributed pits that have reduced pit borders (pit cavities ca.1-2 J.Lm in diameter).In latewood of both species, vessels are sparse or absent, depending on the growth ring, and imperforate tracheary elements are clearly tracheids.Fiber-tracheids in early wood plus tracheids in latewood have also been reported in Carpinus, Corylus, and Ostrya of the Corylaceae or Betulaceae (Metcalfe and Chalk 1950) .Note should be taken that the concept of vascular tracheids, in which extremely narrow vessels grade into tracheids (which are, in effect, vessels lacking perforation plates) in the last few layers of latewood, is quite a different concept.
Mean lengths of imperforate tracheary elements in the family (Table I, column 7) range from 493 J.Lm to 1602 J.Lm , a considerable range.The species with the shortest mean imperforate tracheary element length are Myrica gale (60 I J.Lm) and M. hartwegii (493 J.Lm) .
Mean wall thickness of imperforate tracheary elements follows a pattern similar to the lengths.The means for Myrica gale and M. hartwegii are 1.8 urn and 2.3 urn, respectively, whereas the range in species means for the three other genera is from 2.4 urn to 5.5 urn (Fig. 6,29).In species with moderately to strongly demarcated growth rings, differences in wall thickness between earlywood and latewood occur.For example, in Comptonia (Fig. 8), earlywood tracheid wall thickness is about 2.3 u.m, whereas latewood wall thickness is about 4.7 urn.The claim by Metcalfe and Chalk (1950) that M. gale has imperforate tracheary element walls thicker than those of other Myricaceae is not confirmed here; the reverse was observed.
Mean pit cavity diameter of tracheids is essentially the same as the pit cavity diameter of lateral wall pits of vessels for any given species of Myricaceae (see data for vessel pits above).Contrary to what one sees in conifers and in many dicotyledon families, pitting is denser on tangential walls of tracheids in Myricaceae, sparser on radial walls.

Axial Parenchyma
Axial parenchyma in Myricaceae is basically diffuse; diffuse parenchyma can be found in all species (Table 1, column 9).In most species, diffuse-in-aggregates (tangential aggregates of two to about five cells) is commonly present in addition (Fig. 29).The type "diffuse" implies random distribution of parenchyma throughout the fascicular secondary xylem.Diffuse-inaggregates is recognized because it is a departure from randomness.In Comptonia peregrina, Morella cerifera, M. faya, M. quercifolia, and M. rubra, axial parenchyma cells adjacent to vessels seem a little more numerous than randomness would dictate, so scanty vasicentric parenchyma is said to be present in these species.
In addition, Canacomyrica possesses axial parenchyma distributions not seen in the other Myricaceae: apotracheal bands and apotracheal groupings circular in transection.The latter type is infrequent enough in dicotyledons so that it has not been recognized with a term, so "apotracheal columns" is proposed here for groups composed of about five to ten cells as seen in transection.Longisections of these types are seen in Fig. 2 (upper left, crystal-containing cells) and Fig. 3 (upper left).Arrows indicate apotracheal strands in Fig. 4, and in Fig. 5, a strip of apotracheal parenchyma runs down the center of the photograph (about 14 the width of the photograph).
Axial parenchyma strands, as seen in longisection, range from 3-8 cells.Varying numbers of these cells may be subdivided into cuboidal crystal-containing cells (Fig. 21).Thus, a strand might consist of three or four undivided cells plus the equivalent of two more such cells subdivided into cuboidal crystal-containing cells ("chambered crystals").No subdivision into crystal-containing cells was observed in the two species of Myrica s.s. or in Comptonia, and the lack of such crystal strands is apparently a diagnostic feature of these two genera.Crystal-containing strands of cuboidal cells were observed in Canacomyrica and in all of the species of Morella, although these strands are scarce in some of the Morella species and abundant in others.The nature of crystals in these strands is discussed further below under "Crystals."

Vascular Rays
As shown in Table I, column 10, multi seriate rays are more abundant than uniseriate rays, or approximately equal in abundance in Canacomyrica (Fig. 5) and in Morella (Fig. II,13,15,17,27,31,33,39,and 4 L).Biseriate rays are considered multiseriate here for purposes of comparisons.In contrast to those two genera, uniseriate rays are more common than multiseriate rays in Comptonia (Fig. 9) and Myrica S.s. (Fig. 43,45).The two species of Myrica S.s.differ, however, in that multiseriate rays are scarce and narrow in M. gale (Fig. 43), in which the mean multiseriate ray width (at the widest point) is 2.1 cells (Table I, column II).Multiseriate rays are more common in M. hartwegii (Fig. 45), in which mean multiseriate ray width is 4.1 cells.However, this comparison may be misleading because the specimen of M. gale was from a narrower stem-the wood of a more juvenile character, therefore-than that of M. hartwegii.Rays tend to increase in width as woody stems increase in diameter (Barghoorn 1940).The widest multiseriate rays in the family were observed in Canacomyrica (Fig. 5) and Morella salicifolia (Table I, column II).In Morella, greater width of multiseriate rays was observed in M. inodora (Fig. 15), M. javanica (Fig. 26) and M. nagi (Fig. 17), but narrower rays characterize M. californica (Fig. I I), M. ce rifera, M. domingana, M. faya (Fig. 13), M. kraussiana (Fig. 3 L), M. pubescens (Fig. 33), M. quercifolia (Fig. 41), and M. rubra (Fig. 39).Of these, narrower rays might be related to small sample diameter in M. kraussiana and M. quercifolia, although not in the other species.
With respect to ray histology, the rays of most Myricaceae qualify as Heterogeneous Type IIA, in which multiseriate ray tips are typically more than a single cell tall (Fig. 45; most rays of Fig. 13), or Heterogeneous Type IIB, in which multi seriate ray tips are usually a single cell tall (Kribs 1935; Carlquist 200 I).Rays of the latter type can be seen in all species of Morella, and are shown here in Fig. 27.Where multiseriate rays are very scarce (Fig. 43), one can say that the ray type is transitional between Heterogeneous Type II and Heterogeneous Type III, and closer to the latter (in which all rays are uniseriate), These assignments for rays in Myricaceae agree with those of Metcalfe and Chalk (1950).However, Metcalfe and Chalk (1950) state that in Myricaceae, the multiseriate portions of multiseriate rays are "sometimes with sheath [= upright] cells."Such sheath cells are, in fact, very scarce in most species, but are relatively common in rays of Canacomyrica (Fig. 5), which also has elongate uniseriate tips on multiseriate rays and therefore is referable to Heterogeneous Type I of Kribs (1935).In Myricaceae as a whole, the multiseriate portions of multi seriate rays consist almost wholly of procumbent cells (except in Canacomyricat and the uniseriate tips of multiseriate rays consist of upright cells.Uniseriate rays consist of upright cells (Fig. 2, lower 213 of photograph).Occasional square cells may be observed in radial sections, but they are much less common than either upright or procumbent cells (Fig. 3).
Mean height of multiseriate rays in Myricaceae is relatively low and relatively uniform, 379 urn to 784 urn (Table I, column 12) except in Canacomyrica (Fig. 5), which has a mean ray height of 1598 urn .The mean multiseriate ray height for the family as a whole is 539 I-Lm, and this figure is reflected visually in the rays shown in Fig. 9, II, 13, 15, 17,31,33,39, 41, and 45. Metcalfe and Chalk (1950) report ray height exceeding 1000 urn for M. nagi, but these rays must be infrequent and probably in wood from the base of a relatively large tree.Metcalfe and Chalk (1950) did not study wood of Canacomyrlca.
So-called aggregate rays (clustering of uniseriate or biseriate rays into larger units: IAWA Committee on Nomenclature 1964) are reported most commonly in Fagales, but occasionally in other families (Carlquist 200 I).Aggregate rays have not been reported in Myricaceae, apparently.A phenomenon that some might refer to this phenomenon is visible in some Myricaceae.In transection, one can see multi seriate rays that abruptly break into uniseriate or biseriate rays; these slender rays, in large stems, become wider toward the outside of the stem, in agreement with the increase in ray width with age noted by Barghoorn (1940) .The association of uniseriate rays into groupings (perhaps represented in Fig .17, lower left) decreases rapidly with age, and certainly does not increase as the stems increase in diameter, as the term "aggregate," indicating a kind of "coming together," suggests.The terminology for the rapid breakup of wide primary rays as secondary growth begins in Myricaceae is problematical, therefore.The idea that aggregate rays are large multiseriate rays in the process of disintegration is a generalization offered by Bailey and Sinnott (1914) and Barghoorn (1941), but the reverse (progressive clustering or union of uniseriate rays with age of stem) has been demonstrated repeatedly (see Carlquist 200 I) .The existence of multi seriate rays near the pith of Myr-icaceae might be considered an example of disintegration of multiseriate rays to form aggregate rays, but the abruptness of the process in Myricaceae probably disqualifies them from this designation .
Uniseriate ray height is given in Table I , column 13, As with multiseriate rays, relatively great mean uniseriate ray height was observed in the material of Canacomyrica, but uniseriate rays in Morella, Myrica S.S ., and especially Comptonia are shorter.
Ray cells have secondary walls in all Myricaceae.Bordered pits are common in most species, especially on tangential walls, but simple pits are more common in some species.Bordered pits are shown very clearly in sectional view in rays of Morella inodora (Fig. 19).

Crystals
Rhomboidal crystals occur abundantly in cuboidal cells of the axial parenchyma strands of all types of axial parenchyma groupings of Canacomyrica (Fig. 2,  7) .These cuboidal crystal-containing cells ("chambered crystals") contain rhomboidal crystals covered with a layer of secondary wall material ("encapsulated crystals").Similar strands of cuboidal crystal-containing cells occur in some sheathing cells on multiseriate portions of multi seriate rays of Canacomyrica.Crystals are also abundant in the central portions of multiseriate rays of Canacomyrica, in which they occur in radial series of procumbent cells (Fig. 3).Many such crystals occur in vertically or horizontally subdivided procumbent ray cells.Crystals are so abundant in wood of Canacomyrica that wood sectioning inevitably produces imperfect results.
Outside of Canacomyrica, crystals in rays were observed only in Morella quercifolia (Fig .41) and M. rubra.
All species of Morella were observed to have chambered crystals in axial parenchyma strands.Some of these crystals are so heavily covered (encapsulated) with secondary wall material that virtually no lumen space remains in the cell.This covering can be seen on the lower two crystals in Fig .21.The crystals in Fig. 21 are more complicated in shape than simple rhomboids.Crystals in axial parenchyma can be seen in transections (e.g., Fig. 28).Crystal strands are common in some species of Morella, rare in others.However, a careful search revealed at least a few chambered crystals in all species of Morella.Complete absence of chambered crystals from the genus cannot be claimed in any species of Morella in the present study.
No crystals were observed in wood of Comptonia or Myrica s.s.

Starch and Resinlike Deposits
Starch grains were observed in axial parenchyma of Morella californica and M. nagi, and in the axial pa-renchyma and rays of M. salicifolia.Doubtless starch is more widely present in wood of the family, but was not observed because some methods of preservation tend to promote loss of starch (e.g ., bacterial action during slow drying).Some objects I interpret as starch remnants were observed in several species of Myricaceae.

RESULTS : BARK
In Comptonia bark, there is a band of sclereids separating outer from inner cortex (not shown in Fig. 8), Scattered druse-bearing parenchyma cells and strands of fibers occur in secondary phloem of Comptonia (Fig. 8,top).
In Myrica gale (Fig. 42), druses are present in the outer cortex.A layer of sclereids is present in the inner cortex .In M. hartwegii (Fig. 44), numerous druses occur in the outer cortex along with a few fibers, but no sclerenchyma layer was observed .
In Morella quercifolia, rhomboidal crystals occur in idioblasts in the outer cortex and a sclerenchyma layer encircles the stem in the central cortex.Fiber bands and rhomboidal crystal-containing idioblasts are present in secondar y phloem.
Where the band of sclereids (forming a cylinder in three dimensions) occurs in the cortex of Myricaceae, it is continuous around the stem regardless of the age of the sample (at least in the present study).Breaks inevitably occur in the cylindrical band of sclereids as the xylem cylinder increases in diameter, but as in other dicotyledons with such a sclerenchyma cylinder in bark, parenchyma cell s adjacent to the sclerenchyma band doubtless divide, intrude into the break, and become converted into sclereids.
In all species of Myricaceae for which bark was available, periderm is present outside of the cortex and consists of cells much narrower radially than tangentially.There is a tendency for phellem to exfoliate as thin sheets.The stem specimens of species for which bark was available were relatively sm all in diameter, so persistence of the cortex by means of tangential stretching and radial subdivision of cortical cells was observed.In bark of older stems, loss of cortex is to be expected.

SYSTEMATIC AND PHYLOGENETIC CONCLUSIONS
The four genera of Myricaceae are so different on the basis of wood anatomy that diagnostic summaries, given below, can be developed.These summaries, however, could eventually be expanded more precisely in terms of quantitative features if wood samples of similar degrees of maturity were available.Some quantitative estimates are included in these summaries in the form of terms such as "short" and "intermediate," and these designations can be translated into more precise numerical terms by reference to Table 1 The above comparisons support the classification of Elias (1971) who, how e ver, recognized Morella and Myrica s.s. as subge nera rather than as genera; the diagnostic features above support recognition of these as genera.The ray type Heterogeneous Type I and the re latively large number of bars per perforation plate found in Canacomyrica, a genus not considered by Elias, are features that are primiti ve within the family .Aggregation of axial parenchyma into bands and columns may, however, be an autapornorphy in Canacomyrica .These features are compatible with a myricaceous placement for Canacomyrica.The similarity between Canacomyrica and Morella with respect to crystals in rays and chambered c rys ta ls in ax ia l parenchyma is striking .The absence of crystals in Comptonia and Myrica s.s.contrasts with their prese nce in Canacom yrica and Morella.
Presence of sc alariform perforation plates a nd diffuse axi al parench yma (and vari ations on diffuse parenchyma) are doubtles s symples io morp hies for the family, although reduction in number of bars per perforation plate (most notable in Comptonia) is a probable apomorphy.The presence of dark-colored deposits in the wood, although not a striki ng feature, unites My rica cea e ; Heterogeneous Type IIA or lIB rays are c haracter istic of all genera but Can acomyrica.Trac he id dimorphism and the thin-walled nature of tracheary elements in Myrica s.s., in whi ch earl ywood contains numerous vessels but latewood con sists of tracheids into whi ch few or no vessels are interspersed, are dist inct ive features that se para te Morella from Myrica s.s. and are likely an autapomorphy in Myrica s.s.Other features of Myrica s.s .that differentiate it from Morella include greater grouping of vessels, shorter vessel ele ments, more numerous bars per perfor ation plate, pre sen ce of tracheids in latewood but fiber-tracheids in earlywood, predominance of uni se riate ray s, and absence of crystals.Th ese two genera ar e a mply justified on the basis of wood features.Comptonia is di stinctive with respect to its large number of narrow vessels, change in tracheid wall thickness w ith respect to posit ion in growth rings, and ex-tremely low mean number of bars per perforation plate.Thus, the recognition of four genera is supported.Sampling of Morella here is insuffi cient to judge whether subgenera proposed within Morella (see Eli as 1971 ) receive any support from wood anatomy, however.
Inclusion of Myri caceae in Fagales (near Juglandaceae , Casu arin aceae , and Betulaceae) is proposed by rec ent phyl et ic stud ies that utilize DNA ana lysis (Manos and Steele 1997 ;Soltis et al. 2000) Myricaceae contain more primitive character states than any other fagalean family mentioned above except for Ticodendraceae, which are more primitive only in the scalariform perforation plates (which have more numerous bars and which retain pit membrane remnants in many perforations).Thus, Myricaceae are rich in symplesiomorphies for Fagales.Most of the features listed above are generally considered by wood anatomists as indicative of primitive wood structure, and are thus symplesiomorphies, which are not considered by cladists as indicative of relationship.The distinctive modes of crystal occurrence shared by Myricaceae and Juglandaceae probably represent synapomorphies, underlining the alliance of the two families indicated in recent DNA analyses.
The retention of tracheids in many of the families of Fagales listed above is interesting because the feature occurs in Rosaceae also (Metcalfe and Chalk 1950), and thus is a symplesiomorphy including eurosids other than Fagales.From an ancestry based on an all-tracheid background of imperforate tracheary elements in secondary xylem, instances of vasicentric tracheids have developed in Fagaceae (vasicentric tracheids plus libriform fibers in all genera except Fagus), Casuarina (vasicentric tracheids plus fiber-tracheids in a few species), and chaparral species of Prunus (e.g., P. ilicifolia, P. lyoni) in Rosaceae (Carlquist 1985;Carlquist and Hoekrnan 1985), These are interesting instances of tracheid dimorphism (Carlquist 1988).Thus, libriform fibers (and in Casuarina, fibertracheids) are apomorphic in the species listed above as instances of vasicentric tracheid presence.As will be shown in a study in progress, this pathway is more common than innovation of tracheids in a group with an axial wood background of Iibriform fibers as a basic condition, although that pathway does occur (e.g ., Rosmarinus of Lamiaceae, S. Carlquist, unpublished data).
The FN ratio (see Table I, column 14 for values and definition) can be seen as an index of phyletic advancement, because primitive imperforate tracheary elements (tracheids) are only a little longer than the vessel elements they accompany.Specialized imperforate tracheary elements (Iibriform fibers) are much longer than the vessel elements they accompany (Carlquist 1975).Judged by this interpretation, Myricaceae have rather primitive wood, an interpretation reinforced by the presence in the family of features I, 2, 4, and 5 cited earlier in this section.The F/V ratio is not.however, a precise measure of phyletic specialization, because other factors can be involved in this ratio (e.g., succulence).

ECOLOGICAL CONCLUSIONS
The degree of vessel grouping in Myricaceae is very low (Table I, column 1).The only exceptions of note are in the two species of Myrica s.s., especially M. hartwegii.Myricaceae are one more clear demonstration of the principle that in those dicotyledon families with tracheids in wood, vessel grouping is minimal, whereas more appreciable grouping occurs (to a progressively greater degree with greater xeromorphy) in woods with fiber-tracheids or libriform fibers around vessels (Carlquist 1984).The early wood of Myrica gale and of M. hartwegii contains fiber-tracheids, so that grouping of vessels is an advantageous conductive safety device in earlywood.Latewood in these two species has tracheids, and thus, vessel grouping does not occur in latewood.Indeed, there are few vessels in latewood of these species: the tracheid is the conductive cell type with optimal safety (resistance to embolism formation and to spread of embolisms).This is the physiological significance of the Type V growth ring (Carlquist 1980(Carlquist , 2001) ) .
The range in vessel density might be expected to be approximately inverse to vessel diameter because of packing considerations.Deviations from such an inverse relationship do occur, however, as in lianas, in which greater vessel density probably is related to the relative paucity of mechanical tissue in this growth form, in which stems are not self-supporting.In Myricaceae, notable departure from the inverse relationship occurs in Comptonia peregrina.in which vessels are more numerous per mrn? than would be expected (e.g., the mean vessel diameter in C. peregrina is the same as that in M. gale, which has about a quarter as many vessels per mm-) .The elevated vessel density in C. peregrina may relate to its habitats, which are probably among the driest for Myricaceae.Although tracheids in Comptonia offer conductive safety, presence of a very high number of vessels offers another form of conductive safety by providing a redundancy of vessels that would insure that some of them (presumably mostly latewood vessels) would resist embolism in times of drought and frost.
The family Myricaceae as a whole ch aracterizes moist habitats, including tho se whe re water stands for prolonged periods, resulting in low nitrat e conditions.Nitrogen fixation in roots of Myrica ceae by actinomycetes has been repeatedly ob served (Van Ryssen et al. 1970;Turner and Vitousek 1987;Sprent et al. 1978) .Certainly Myricaceae have wood features that qualify as mesi c.The Mesomorphy Ratio (see Table I, column 15, for values and definition) value for all studied Myricaceae, averaged , is l250.Th is figure is higher than tho se for all but a few southern Californian plants (Carlqui st and Ho ekman 1985).Because the habitats most commonly occupied by Myricaceae (sw a mps wi th fluctuation of wa ter level ; strea msides; slopes or flats w ith steady subsurface wate r availab ility) are not easy to defin e in terms of rainfall, but do exemplify various kinds of moisture availability, compari son of the rather high Mesomorphy Ratio values of species of Myricaceae with those in oth er habitats is d ifficult.Su ch high values, however, would be expected in tropical cloud fores t shru bs and tree s (e .g., moist for est species of the Hawaiian genu s Dubautia, Carlquist 1998).
AC KNOWLEDGMENTS Dr. Regi s M iller kindly pro vided samples from the xylaria of the Forest Products Laboratory, Madison, Wisconsin, and Mr. Stanley Yankowski mad e several sa mples available from the xylarium of the U.S. National Mu seum of Natural History.Much of the work was done using the facilities of the Ran cho Santa Ana Botanic Garden , which is gra tefully acknowledged; the remainder of the work was accom plished at Santa Barbara Botanic Garden.Samples of Morella from South Afri ca were coll ected with the aid of a grant from the American Ph ilosophical Society.
Fig.1-5.Canacomyrica monri cola.-J.Fruit o n inflorescence ax is, bracteole at base and a nthe rs at ape x are di stinguish ed by their coverings of pale g land ula r trichornes ; s tig mas (da rk gra y ) at summ it o f fruit; note pointed struc tures , which may be cal yx teeth, be low anthers.-2-3.Rad ial sec tio ns of wood .-2 .Strands of ax ia l parench ym a bearing rh ombo idal c rysta ls (a bove) a nd upri ght cells of un iseriate ra y (be low).-3.Radi al file s o f c rysta l-bearing cells in mult iseriat e ray (a rro ws) , se pa rated from eac h othe r by ce lls that lack c rys ta ls .---4 .Tran sect ion o f wo od , narrower lat e wo od ves se ls near bo uo rn.v-o .Ta nge ntia l sec tio n of wood ; uni se riate rays ha ve up right cell s, mult iseriate rays ar e co mpo se d o f pr ocumbent ce lls with some upri ght sheathing ce lls .Fig .I , sca le bar = J mm ; Fig. 2, sc ale bar at top, d ivi sions = 10 urn : Fig. 3-5, scal e bar at top o f Fig. 3, divisi on s = 10 urn.

Fig . 22 -
Fig .22 -25 .SEM photograph s o f radial wood sec tio ns of Morella.-22 .M. krau ssiana.Perforat ion plate , with relativel y narr ow .numerou s bar s.-23-25.M. l1agi .-23.Perforation plate with bars o f typi cal th ickn es s: bordered pits o f trach e ids are con spi cu ou s to left and right of the vessel.-24 .Perforat ion plate with rel ati vel y few , wide bar s; prom inent borde red pits of Irach eid s 10 the right of the vesse l.-25.Porti on s o f two trach e ids. to sho w cha rac ter istic bord er diameter s and pit a perture shapes .Fig. 22-24, sc a les = 10 urn : Fig .25, sca le = 2 urn.

Fig. 26
Fig. 26-29.Wood sec tions of Mor ella .-26-28.M. javanica.-26.Transe ction : growth rings absent , vessels notabl y wide.-27 .Tangenti al sec tion; wide as well as narrow multi seriate rays are presen t: both multiseriate and uniseriate rays co ntain dark-stainin g dep osits.-28.Transec tion.showing thick-w alled trach eid s in which depo sits o f dark -staining mater ial outline the bord ered nature of the pits: a pair of rhomboid crys tals.ju st above cente r. and .(0 right a nd left of the c rysta l-bearing ce ll.axial paren chyma ce lls.-29.M. saticifo lia: transection: axial paren chyma is diffu se and. a little below cente r. a band o f diffuse-in-aggre gates axial parenchyma is present : dark dep osits in rays and in a few of the axial pare nchyma cell s.Fig. 26.27 .scale above Fig.3: Fig. 28.scale above Fig.6: Fig. 29.sca le above Fig. 2.