Wood Anatomy of Hydrophyllaceae. I. Eriodictyon

Hydrophyllaceae is a characteristically herbaceous family: only Eriodictyon Benth.. Wigandia H.B.K.. and the monotypic Turricula Macbr. can be considered woody (stems attaining a diameter of more than 1 cm). Are plants such as these relicts from a woody ancestry, or is the family secondarily woody? The nature of wood anatomy offers some features useful in the attempt to answer this question. Wood anatomy is also applicable to analysis of the curious growth form of Eriodictyon. Unlike most shrubs. Eriodictyon is characterized by a welldeveloped root system from which aerial innovations of various height and duration arise. Such a growth form is ideal for resisting drought and fire in the chaparral and chaparrallike areas where Eriodictyon grows. The majority of Eriodictyon plants (using that term to denote an aerial stem) in any given locality may ultimately result from root innovations rather than events of seed germination. Indeed, attempts to grow Eriodictyon from seed have been unsuccessful (Everett 1957). Because roots are persistent, large, and of ecological significance in the genus, wood of them has been analyzed for three species (Table 1). For two other species, a portion termed underground stem has been studied. In E. trichocalyx Heller subsp. lanatum (Brand) Munz the underground stem immediately subtending the aboveground stem was collected. In E. traskiac Eastw. subsp. smithii Munz. the underground stem was similar, but further buried beneath the ground by the fallen rubble of diatomaceous earth typical of its habitat. Shrubs of Eriodictyon range in size from a maximum of 4 m in E. altissimum P. V. Wells (Wells 1962) or 3 m in E. crassifolium Benth. (Munz 1959) to shrubs barely reaching VA m in some subalpine populations of E. californicum (H. & A.)Torr. and E. trichocalyx subsp. trichocalyx.


INTRODUCTION
Hydrophyllaceae is a characteristically herbaceous family; only Eriodictyon Benth., Wigandia H.B.K., and the monotypic Turricula Macbr. can be considered woody (stems attaining a diameter of more than 1 em). Are plants such as these relicts from a woody ancestry, or is the family secondarily woody? The nature of wood anatomy offers some features useful in the attempt to answer this question.
Wood anatomy is also applicable to analysis of the curious growth form of Eriodictyon. Unlike most shrubs, Eriodictyon is characterized by a welldeveloped root system from which aerial innovations of various height and duration arise. Such a growth form is ideal for resisting drought and fire in the chaparral and chaparrallike areas where Eriodictyon grows. The majority of Eriodictyon plants (using that term to denote an aerial stem) in any given locality may ultimately result from root innovations rather than events of seed germination. Indeed, attempts to grow Eriodictyon from seed have been unsuccessful (Everett 1957). Because roots are persistent, large, and of ecological significance in the genus, wood of them has been analyzed for three species (Table 1). For two other species, a portion termed underground stem has been studied. In E. trichocalyx Heller subsp. lanatum (Brand) Munz the underground stem immediately subtending the aboveground stem was collected. In E. traskiae Eastw. subsp. smithii Munz, the underground stem was similar, but further buried beneath the ground by the fallen rubble of diatomaceous earth typical of its habitat. Shrubs of Eriodictyon range in size from a maximum of 4 m in E. a/tissimum P. V. Wells (Wells 1962) or 3 m in E. crass~folium Benth. (Munz 1959) to shrubs barely reaching 1 /4 m in some subalpine populations of E. cal(fornicum (H. & A.)Torr. and E. trichocalyx subsp. trichocalyx.
Wood anatomy is also worthy of investigation with respect to the ecological range of the genus. Eriodictyon is a remarkably characteristic shrub of several regions of California; nine of the 10 species occur in the state. All of the species currently recognized as valid according to recent floras (e.g., Munz 1959, with the addition of E. altissimum; Wiggins 1980) are included in the present study, and the two subspecies recognized for two of the species are represented (Table 1). Dubiously distinct varieties of E. crassifolium have not been included. Eriodictyon sessilifolium Greene is restricted to coastal regions of northern Baja California, Mexico (Wiggins 1980). Of the nine species indigenous to California, some range into other states. Eriodictyon angustzfolium Nutt. extends from the eastern Mojave Desert of California to Utah, Arizona, and Baja California (Munz 1959), and could be said to occupy drier and (in summer) hotter sites than do the other species. Eriodictyon californicum occurs in the Sierra Nevada and northern Coast Ranges of California, from which it extends into Oregon. Eriodictyon trichocalyx subsp. lanatum grows both in the southernmost portion of California and in the adjacent parts of Baja California. Only E. traskiae subsp. traskiae occurs on an offshore island, Santa Catalina Island. None of the habitats occupied by Eriodictyon could be said to be mesic. While some of the habitats occupied by the genus have moderated temperatures by virtue of proximity to the seacoast (habitats of E. altissimum, E. capitatum Eastw., E. sessilifolium, and E. traskiae), other habitats (those of E. californicum and E. trichocalyx) experience considerable frost.
Although Eriodictyon grows in dry places, the vegetative aspect of most species is not extremely xeromorphic. To be sure, the narrow leaves of E. angustifolium suggest adaptation to excessive heat loading ofleaftissue and water loss from transpiration by virtue of adoption of a narrow leaf shape. The leaves of the coastal E. altissimum are also notably narrow, however. Because leaves of Eriodictyon are not truly microphyllous, they must have mechanisms other than size and shape for minimizing heat loading and water loss. The surface characteristics of indument and varnish, although they may well function primarily against herbivore and insect predation, may have some function in protection from heat and drought. The wood anatomy of Eriodictyon may well be an important water-management device, but the nature of the foliage doubtless is important and must ultimately be taken into account.
Woods of Hydrophyllaceae are notable in having tracheids bearing full bordered pits rather than libriform fibers with simple pits (Record and Hess 1943;Metcalfe and Chalk 19 50). Some other tubiflorous families of dicotyledons have such tracheary elements, notably Convolvulaceae ' (Metcalfe and Chalk 1950), Dipsacaceae (Carlquist 1982) and Goodeniaceae (Carlquist 1969). However, this allegedly primitive feature is not common in the tubiflorous families of dicotyledons. Do Hydrophyllaceae have other primitive wood features? Wood anatomy is potentially of interest in assessing phylogenetic position, degree of specialization, and shift to or from woody habit in Hydrophyllaceae.

MATERIAL AND METHODS
The taxa studied and the specimens documenting them are shown in Table  L Portions ofwood from the plant parts indicated in column 1 were mostly < Legend for numbered columns: I, plant portion (R =root, S =stem of maximal diameter near base of plant, SS =stem of small plant in unfavorable locality, SM = stem of medium-sized plant in intermediate locality, SL = stem of large-sized plant in favorable locality, US = underground stem). 2, mean vessel diameter, I'm. 3, mean number of vessels per mm 2 • 4, mean vessel-element length, I'm. S, mean number of vessels per group. 6, mean vessel wall thickness, I'm. 7, mean tracheid diameter at widest point, I'm. 8, mean tracheid length, I'm. 9, mean tracheid wall thickness, I'm. 10, mean width ofmultiseriate rays at widest point, cells. II, mean height multiseriate rays, I'm. 12, mean height uniseriate rays, I'm. 13, Growth-ring type according to scheme of Carlquist (1980). 14, Index "Vulnerability" (Carlquist 1977). IS, Index "Mesomorphy" (Carlquist 1977). preserved in 50% ethyl alcohol. The use of pickled wood specimens revealed few details not also visible in dried woods other than presence of nuclei in axial and ray parenchyma. Woods were sectioned on a sliding microtome. In a few, hardness made use ofKukachka's (1977) ethylenediamine softening method desirable. Sections were stained in a safranin series, as were macerations, which were prepared with Jeffrey's Fluid. Quantitative data in Table  l are based on 25 measurements for each feature.
The majority of sections and data were prepared by Vincent M. Eckhart. Dr. David C. Michener contributed wood of four of the taxa, as shown in Table 1. Dr. Leila Schultz kindly collected the material of E. sessilifo!ium. Two of the species are in localities of limited access. Of these, Dr. Gary Hannan provided material of E. capitatum, while Dr. D. D. Dingus of California State Polytechnic University at San Luis Obispo kindly collected stems of E. altissimum. He collected three stems of E. altissimum (see Table  1, column 1). These stems were 1.5, 3.2, and 4.3 em in diameter, respectively. These correspond to sites judged by Dr. Dingus to be dry, less dry, and moderately moist, respectively. No herbarium specimens were preserved for the collection of E. altissimum. Herbarium specimens documenting the other taxa of Eriodictyon are located in the herbarium of Rancho Santa Ana Botanic Garden, with the exception of E. capitatum, the specimen of which is located at University of California, Santa Barbara.

ANATOMICAL DESCRIPTIONS
Growth rings.-The types of growth rings found in Eriodictyon are listed in Table 1. The types are not, of course, sharply demarcated from each other and intergradation among them is to be expected. Minimal growth ring activity is shown by E. sessilifolium ( Fig. 9), in which earlywood vessels are a little wider than vessels in latewood (Type IB), although not greatly so. In certain taxa of Eriodictyon, growth rings are not strongly demarcated, but tracheids as well as vessels are wider in earlywood than in latewood (Type ID: E. capitatum roots, Fig. 13). If growth rings are sharply delimited, with wider vessels and wider tracheids in earlywood, Type VD is said to exist (most Eriodictyon collections: e.g., E. altissimum, Fig. 5; E. traskiae subsp. traskiae, Fig. 15). In a few, parenchyma tends to substitute for tracheids in earlywood, and thus Type IC can be said to be present (e.g., E. angustifolium roots, Fig. 2, 3). There is in some collections a conspicuous tendency for vessels not to be largest in the earliest-formed earlywood, but to increase after the beginning of the growth ring. Designated Type X (Carlquist 1980), such growth rings are shown here for stems of E. angustifolium ( Fig. 1) and, less conspicuously, stems of E. captitatum (Fig. 11). When a Type X growth ring occurs, one presumes that the cambium becomes active before the period of maximal water flow begins. Thus, smaller vessels would be produced as the wet winter months begin, but larger vessels would follow during the wet, but warmer, spring months. This presumption was confirmed chronologically in the stems of E. traskiae subsp. smithii, collected on December 22, 1982, after appreciable winter rainfall had already occurred, and some earlywood containing smaller vessels had been produced at the periphery of the stem.
Vessel elements.-Vessels are round in shape as seen in transection ( Fig.  1, 2 , 3, 5, 9, 11, 13, 15), compressed when in contact with other vessels but never truly angular. Record and Hess (1943) claim that vessels in Hydrophyllaceae are solitary (their descriptions based on Eriodictyon plus Wigandia). Although vessels are very nearly solitary in Wigandia, various degrees of grouping are exhibited by Eriodictyon (Table 1, column 5). Figures below 1.20 for this feature certainly indicate a predominance of solitary vessels, but half of the collections lie above this value. The larger number of vessels per group tend to occur in the species from drier and colder areas: E. altissimum, E. angustifolium, E. californicum, and E. trichocalyx. One could say, however, that more numerous vessels per group is an inevitable byproduct of having more numerous vessels per mm 2 of transection. That correlation is generally close, and one notes from comparing columns 3 and 5 in Table 1 that roots tend to have both few vessels per mm 2 and few vessels per group compared to stems. The correlation between vessels per mm 2 and vessels per group is not entirely linear, however; if it were, there would be no value in presenting both measurements. Of the two measurements, vessels per mm 2 is probably more significant, because it is a direct expression of the indicator of redundancy (or lack of it) in the conductive system. Although some groups, such as Asteraceae (Carlquist 1966) show increased numbers of vessels per group with increasing xeromorphy, some families of dicotyledons show little or no such elevation in these values.
Mean vessel-element diameters (Table 1, column 2) in Eriodictyon are typical of xeromorphic woods in the case of stems, but the wider vessel elements in roots are clearly mesomorphic by comparison. Such side-byside comparisons of stems and roots are offered for E. angustifolium (Fig.  1, 2) and E. capitatum (Fig. 11, 13). The taxa with wider vessels tend to have fewer per mm 2 , within limits.
The length of vessel elements in Eriodictyon (Table 1, column 4) is below the median for dicotyledons as a whole (cf. Metcalfe and Chalk 1950, p. xxiv). The average vessel-element length for the 20 collections of Eriodictyon is 385 J.Lm. This is certainly moderately xeromorphic, but longer than the average vessel-element length for Asteraceae as a whole, 235 !lm (Carlquist 1966).
Fibriform vessel elements.-Eriodictyon wood as seen in transection il-  (Fig. 5-7 , 9, 10, scale above Fig. 1; Fig. 8, scale above Fig. 3.) lustrates a range of vessel diameter, so that vessels are difficult to differentiate from parenchyma cells (which have wall thickness approximately the same). In a maceration of Eriodictyon wood one sees some narrow vessels also.
These are fusiform in shape, and the perforation plates are lateral rather than terminal thereby. In length, these narrower vessel elements, which I am terming fibriform here, tend to be longer than the wide vessel elements. That may, in part, account for the fact that mean vessel element length in Hydrophyllaceae is appreciably greater than that of Asteraceae. This greater length for fibriform vessel elements may, in developmental terms, be the product of greater intrusiveness of tips during maturation as compared with wider vessel elements. The vessel-element length of Hydrophyllaceae, however, may at least in part represent a lower degree of phylogenetic specialization than Asteraceae represents.
As seen in radial sections ( Fig. 17-19), fibriform vessel elements can be seen to have very small perforation plates. Indeed, some of these plates have wide borders and look much like exceptionally wide bordered pits, such as the bordered pits seen on lateral walls of vessels. Some ofthe fibriform vessel elements have more than one perforation (Fig. 18, 19). Study of such double perforation plates reveals that most of these do not relate to branching of a vessel, but interconnect only a pair of vessel elements.
The fibriform vessel elements of Eriodictyon might at first glance be regarded as similar to vascular tracheids, although perforated. This interpretation is rejected, and the following reasons are advanced for regarding the fusiform vessel elements as quite different from vascular tracheids.
(1) The fibriform vessel elements do not occur merely at the end of growth rings; they may be found in the most mesomorphic of the Eriodictyon woods, those of roots, where they can be found in association with wider vessels.
(2) The fibriform vessel elements are longer than wide vessel elements, whereas in most instances vascular tracheid lengths agree with the lengths in any given wood sample.
(3) Thus far all known instances of vascular tracheids are reported from woods of great specialization, where libriform fibers are the imperforate element type. If Eriodictyon had vascular tracheids, therefore, one would also expect libriform fibers to be present; none are. The only wood reported in which vascular tracheids do not coexist with libriform fibers is that of Loricaria thuyoides (Carlquist 1961 ), in which all vessels are narrow at most, and in which narrow vessels and vascular tracheids are so abundant that no mechanical cell type is present and thereby a remarkable gymnospermlike condition occurs. (4) The selective value of a vascular tracheid lies in its ability to confine an air embolism within a single conductive cell, so that air embolisms do not spread to other cells. True tracheids have the same selective value. Therefore, one cannot imagine any reason why a plant which already has tracheids as the imperforate tracheary element type should evolve vascular tracheids. In order to have the advantage of tracheids under water stress conditions, a tracheid-bearing plant merely could produce tracheids only in the later portions of a growth ring. In fact, there are various woods (Ephedra; Bruniaceae) in which this happens, and vessels are absent from the latter part of a growth ring. Fusiform vessel elements are more common in Eriodictyon than in Wigandia, and presumably relate to the selective value of narrower vessel elements, more numerous per unit transection, in a plant or plant organ adapted to a highly seasonal climate.
Tracheids.-Hydrophyllaceae are claimed by Record and Hess ( 1943) and Metcalfe and Chalk (1950) to have tracheids ("fibres with bordered pits") rather than fiber-tracheids or libriform fibers. This is confirmed by the present study (Fig. 16). While Hydrophyllaceae are unusual in this regard among tubiflorous families, one can cite Convolvulaceae (Metcalfe and Chalk 1950), Dipsacaceae (Carlquist 1982) and Goodeniaceae (Carlquist 1969) as families in which tracheids are the imperforate element type. Chloanthaceae (Carlquist 1981) is a tubiflorous family in which fiber-tracheids occur. Vestigial borders on pits of imperforate tracheary elements have been reported so far only in a scattering of species in the families Bignoniaceae, Boraginaceae, Loganiaceae, Polemoniaceae, Scrophulariaceae, Solanaceae, and Verbenaceae. Presence of tracheids in a wood is statistically correlated with other primitive features (Metcalfe and Chalk 1950, p. xlv) in dicotyledons as a whole, and it may be regarded as one of several relictual primitive features in Hydrophyllaceae. Tracheids can be claimed to have a strong selective value in the ecological regimes where Eriodictyon exists, however, as mentioned below.
Axial parenchyma. -Record and Hess (1943) described the axial parenchyma of Hydrophyllaceae as having a diffuse distribution, with a tendency toward reticulate patterning, of these cells as seen in transection. That can be confirmed on the basis of the present materials, among which those patterns appear most clearly in Fig. 15. In some collections, such as the roots of E. angustifolium (Fig. 2, 3), early portions of growth rings feature more abundant axial parenchyma, so that apotracheal bands of parenchyma could be said to be present. Roots have parenchyma more abundantly than do stems in Eriodictyon. Axial parenchyma comprises less than a fourth of the axial tissue shown in the stem tangential sections of Fig. 6, 7, 10, and 12, but it comprises more than a third ofthe cells in the root tangential section shown as Fig. 14. Axial parenchyma is formed as strands of four to six cells in Eriodictyon.
Vascular rays.-Ray cells are predominantly procumbent in multiseriate rays of Eriodictyon (Fig. 4, 8). Uniseriate wings are infrequent on multiseriate rays (Fig. 7, 10, 12). Most frequently, the upper and lower tips of multiseriate rays consist of a single cell as seen in tangential section; such cells are most commonly square or procumbent, sometimes erect. Erect cells do not sheathe  (Fig. 11-14, magnification scale above Fig. 1.) multiseriate portions of multiseriate rays in the genus. Uniseriate rays may be composed, in any given section, wholly of erect cells or wholly of procumbent cells or of both cell types. Rays in Eriodictyon correspond to Kribs's (1935) Heterogeneous Type liB except for the marked procumbence of cells in the multiseriate rays and the lack of sheathing erect cells on the multiseriate portions of multiseriate rays.
At earlier stages in ontogeny of rays (nearer the pith), erect cells are more frequent, although not dramatically. However, multiseriate rays are lacking near the pith in Eriodictyon, as shown in E. altissimum (Fig. 6). Multiseriate rays are initiated by radial longitudinal subdivision of ray initials. Such ontogeny can be found in a scattering of dicotyledons, and is not a clue to relationships of Hydrophyllaceae. In the experience of the senior author, origin of rays ontogenetically as uniseriate rays exclusively is much less common in herbaceous groups than in woody ones.
Multiseriate rays in stems of Eriodictyon are characteristically narrow, mostly averaging less that three cells wide (Fig. 7, 10, 12). Multiseriate rays of roots appear to be much larger than those of stems as seen in tangential section. This proves to be true in three respects: size of cells (e.g., compare Figs. 12 and 14); average number of cells wide at widest point (Table 1, column 1 0); and average height (Table 1, column 11 ). The mean height of uniseriate rays in Eriodictyon is not well correlated with height ofmultiseriate rays for any particular sample. The greater size of multiseriate rays in roots as compared to stems can be considered related to water storage. Although all ofthe samples of root wood and many of stem wood in the present study were pickled, storage of starch was noted (as numerous small grains) only in stem rays of E. angustifo!ium.
Perforate ray cells, shown for E. angustifolium in Fig. 4, where observed in most of the samples of Eriodictyon wood. These cells may be related to breakup of rays during increase in diameter of the stem or root.
Other features.-No storying was observed in wood of Eriodictyon. Deposits of resinlike compounds were seen infrequently (center of Fig. 11; parenchyma cells of Fig. 15). However, such deposits are probably not characteristic of particular taxa, and may at least in some cases represent a trauma response.

ECOLOGICAL SUMMARY OF WOOD ANATOMY
Eriodictyon wood seems strongly adapted to dry habitats, based on criteria developed earlier (Carlquist 1975). Eriodictyon wood has simple perforation plates, which permit rapid flow of water at times of high transpiration and water availability. However, tracheids, which prevent the spread of air embolism during episodes of water stress, are the imperforate tracheary element type. Thus one can add Eriodictyon to the list of desert and dry habitat Californian shrubs with tracheids as the imperforate tracheary element type, a list which includes Adenostoma, Cercocarpus, Crossosoma, Ephedra, Fallugia, Forsellesia, Krameria, Larrea, Mortonia, and Purshia (Carlquist 1980). Functionally similar are those woods of this region which have libriform fibers as the imperforate tracheary element type but which produce vascular tracheids at the end of a season; the list of these includes Arctostaphylos and Artemisia (Carlquist 1980) as well as Keckiella (Michener 1981).
Another striking adaptation by Eriodictyon to dry conditions is found in the divergence between root and stem in wood features. The xeromorphic features of stem woods include narrowness of vessels (notably the fibriform vessel elements), high number of per mm 2 oftransej::tion, and possession of growth rings with wider vessels at the beginning (Type VD) or after the beginning (Type X) of the growth ring but narrow vessels elsewhere in the growth ring. The indices for vessel features termed Vulnerability and Mesomorphy have been calculated according to the formula advanced earlier (Carlquist 1977) and presented in Table 1, columns 14 and 15. Stems of Eriodictyon have Vulnerability and Mesomorphy index values in the ranges found in dry land plants. If figures given for an assemblage of desert shrubs given by Carlquist (1975, p. 206) were translated into the indices, V would be 0.08 and M would be 17.9. That Eriodictyon stems lie in the range from V = 0.10 and 0.45 and M = 33 to 197 makes them somewhat higher than the average of desert shrubs in these indices, but in a close order of magnitude.
That Eriodictyon stem wood is xeromorphic but not extremely so may be explained on the basis of the compensatory effect of roots as succulent storage and shoot innovation organs; possibly foliar characteristics have a mediating effect as well. If shoots can die back to roots in severe conditions, then obviously the role of wood anatomy in water management by stems is moderated. Compared to stems, roots have much more axial parenchyma; rays are taller in dimensions, wider in cell number, and are composed of much larger cells. Roots have wider vessels, fewer per mm 2 compared to those of stems. Because Eriodictyon roots have some succulent characteristics, we may compare their wood to that of an assemblage of succulents (Carlquist 1975, p. 206). These succulents have the values V = 1.12, M = 290. Eriodictyon roots have even higher values. Figures for roots from Table  l are V = 2.53, 4.60, and 2.10; M = 1037, 2131, and 788 for the roots studied, respectively. Roots of Eriodictyon do not appear succulent in gross aspect to the extent tubers would, but their texture and hardness is different from what one finds in stems. Perhaps the higher figures in Eriodictyon roots may be related to the fact that these are underground succulent organs, whereas the figures based on an assemblage of succulents were derived from stems.
Pickled material was available for all roots of Eriodictyon studied; this did not reveal appreciable quantities of starch. For the present, therefore, roots of Eriodictyon qualify primarily as water-storage rather than photosynthate-storage organs. Underground stems, as mentioned earlier, were studied for two species of Eriodictyon. Vulnerability and Mesomorphy values for these prove to between intermediate between stem and root values, not surprisingly.
If species of Eriodictyon are compared with each other, some differences in wood anatomy related to habitat appear. Using only stems as the means of comparison, growth rings tend to be absent (E. sessilzfolium) or less marked (E. capitatum, E. traskiae subsp. smithiz) in taxa of southerly and coastal localities. Inland and montane taxa have sharply demarcated growth rings (most notably E. angustifolium, E. californicum, E. trichocalyx). In any given species, roots have much less marked growth rings than do stems.
Marked fluctuations in humidity and temperature may relate to the xeromorphic woods found in E. californicum (Sierra Nevada) and E. trichocalyx (mountains and valleys of southern California). Elevational differences are apparent, although not statistically significant, in the E. trichocalyx plants studied. The E. trichocalyx subsp. trichocalyx collections Carlquist 15649 and 15651 a are from elevations in the San Gabriel Mountains (near Claremont) where snow occurs, and these have lower mesomorphy values than does Carlquist 15650, from Claremont (elevation ca. 400 m) where snow in winter is virtually unknown.
The number of vessels per group is elevated in the collections of Eriodictyon from more extreme localities, but this is very likely an expression of the greater number of vessels per mm 2 in these woods. The occurrence of fibriform vessels, present throughout the genus, deserves mention. Such narrow vessels, some indistinguishable from tracheids in transection, undoubtedly confer a degree of redundancy (and therefore safety under water stress conditions when a proportion of vessels may be disabled).

PHYLOGENETIC SUMMARY OF WOOD ANATOMY
Wood of Hydrophyllaceae is specialized with respect to wood anatomy, as one might expect in a tubiftorous family of dicotyledons. However, wood of Hydrophyllaceae is less specialized than that of most of those families in the following features: tracheids rather than libriform fibers or fiber-tracheids are present as the imperforate tracheary element type; axial parenchyma is diffuse rather than vasicentric or otherwise grouped; rays are Heterogeneous Type liB rather than homogeneous (but specialized beyond typical liB in predominance of erect cells and lack of erect sheathing cells on multiseriate portions of multiseriate rays. The ratio in length between imperforate tracheary elements and vessel elements can be regarded as an index of phylogenetic specialization in di-cotyledons, within limits (Carlquist 197 5). From the data in Table 1, one can easily calculate such ratios. In Eriodictyon they range from a low of 1.43 (E. crassifolium) to 2.62 (E. trichocalyx subsp. lanatum underground stem). The mean ratio of tracheid length to vessel-element length for the 20 collections of Eriodictyon is 1.86. This is moderately, but not highly specialized.
Wood anatomy does have some evidence to offer concerning whether a group is ancestrally herbaceous or woody. Predominance of erect ray cells is conspicuous in groups which seem ancestrally herbaceous (Carlquist 1962). If this is true, Eriodictyon bears no traces of herbaceous ancestry, since cells ofmultiseriate rays are predominantly procumbent, even at or near the outset of secondary growth. A second pertinent feature is the occurrence of uniseriate rays exclusively near the pith in Eriodictyon. During increase in stem diameter, multiseriate rays develop from these by radial longitudinal division of ray initial cells. This phenomenon happens to occur in woody plants typically: one can cite such groups as Bursera (Barghoorn 1941 ), Illicium, and Sarcococca. Genera other than Eriodictyon, which will be covered in a forthcoming paper, will be important in assessing this and other criteria for whether Hydrophyllaceae are primitively herbaceous or woody.