Aliso: a Journal of Systematic and Evolutionary Botany Population Structuring and Patterns of Morphological Variation in Californian Styrax (styracaceae) Population Structuring and Patterns of Morphological Variation in Californian Styrax (styracaceae)

Recent studies of genetic variation within and among populations and phylogenetic estimates have provided evidence bearing on the evolutionary history and taxonomy of Styrax in California (S. redivivus). In this paper, data from these studies are further analyzed and integrated with new data from morphology to gain insight into the nature and taxonomic significance of character variation within this species. Six morphological characters thought to be important in the delimitation of infra-specific taxa within S. redivivus were measured on 52 herbarium specimens and analyzed with Pearson correlations and multivariate methods. Five characters are significantly associated with latitude and three characters are significantly multiply correlated with latitude. Permutation tests show a significant association between isozyme allelic variation and latitude. Principal components analysis of the morphological data does not reveal distinct clusters. The distribution of character variation shows that most characters vary along continuous latitudinal clines, and no character exhibits an evident gap in character states. Although principal coordinates and neighbor-joining analyses of the isozyme data, and discriminant function analysis of the morphological data suggest the presence of two groups within S. redivivus, the sum of evidence does not support the delimitation of infraspecific taxa. A taxonomic treatment of S. redivivus, a distribution map of historical Californian collections, and a key distinguishing S. redivivus from related taxa are presented. The species status of S. redivivus is justified, and implications of the data for conservation are discussed.


INTRODUCTION
The genus Styrax (Styracaceae) comprises about 120 species of trees and shrubs distributed in eastern and southeastern Asia, the New World, and the Mediterranean region. It is characterized by a 3-locular, superior ovary, usually twice the number of stamens as petals, a campanulate, white, usually 5-parted corolla with petals connate at the base, and a linear style. Leaf arrangement is alternate, and the vesture consists of stellate hairs or scales. Styrax is currently divided into Sect. Foveolaria (Ruiz & Pav.) Perkins, 3-to 5ovulate (two species, Cuba and Peru), and Sect. Styrax, 16-to 24-ovulate (remaining species). Perkins (1907) maintained two series within Sect. Styrax: Ser. 1mbricatae Giirke (about 30 species) and Ser. Valvatae Giirke (about 90 species), delimited on the basis of floral aestivation type. Some species of Sect. Valvatae are variable for aestivation (sometimes even within the same individual), suggesting that the two series are not monophyletic (Steenis 1932).
In far western North America, Styrax is represented by a single species from Ser. 1mbricatae occurring from Shasta County to San Diego County, California 1 Current address: Department of Botany, Duke University, Durham, North Carolina 27708-0339 ( Fig. 1). Californian Styrax was originally described by Torrey (1851) as representing a new genus with one species, Darlingtonia rediviva. Torrey (1853), upon obtaining good flowering specimens of the plant, transferred the species to the genus Styrax as S. californicum, and applied the name Darlingtonia to the genus of pitcher plants, whence it has been conserved. Wheeler (1945) realized Torrey's mistake as to the correct species name and made the combination S. rediviva (Torr.). Howard (1974) recommended that Styrax be treated as neuter, but Nicolson and Steyskal (1976) have provided a detailed argument for the use of the masculine gender. Following their opinion, I have used the masculine ending for Styrax specific epithets here and elsewhere (e.g., Fritsch 1996).
From the time of Perkins's (1907) worldwide treatment of Styrax, the Californian entity has generally been regarded as conspecific with S. officinalis L. from the Mediterranean region (southern Greece to Turkey, south to Israel). Perkins (1907) could find no basis upon which to distinguish Californian and Mediterranean material, and subsumed S. californicum ( = S. redivivus) under S. officinalis. Subsequent workers (e.g., Rehder 1915;Munz and Johnston 1924;Copeland 1938;Gonsoulin 1974;Thome 1978;Murray 1982;Shevock 1993) have treated the Californian rna- terial as one or two infraspecific taxa of S. officinalis. As such, S. officinalis constitutes a remarkable intercontinental disjunction, and has played a role in phytogeographic models of Tertiary Laurasia (e.g., Axelrod 1975;Raven and Axelrod 1978). Nonetheless, Thorne (1978) doubted whether Californian and Mediterranean populations are conspecific and stressed the need for a thorough biosystematic study of the group. Californian Styrax reportedly differs from the Mediterranean entity in a number of characters: rusty brown versus tawny-to white-stellate stalked trichomes on new twigs, petioles, and abaxial leaf surfaces, fewerflowered racemes, larger flowers, thickened pedicels, a longer staminal tube, a more compressed style, and more pronounced stigma lobes (Rehder 1915;Munz and Johnston 1924;Gonsoulin 1974). Gonsoulin (1974) considered these differences to be minor and inconstant.
Opinion has differed regarding the utility of recognizing infraspecific taxa within Californian Styrax and what rank these should be given. Gonsoulin's (1974) treatment followed Eastwood ( 1906) and Munz and Johnston (1924) Thorne (1978) treated the varieties as subspecies. Variety fulvescens reportedly differs from var. californicus in having suborbicular or ovate-orbicular versus ALISO broadly ovate laminae, broader and cordate at the base; heavily pubescent versus glabrous or puberulent abaxial leaf surfaces and calices; and a general prevalence of rufous hairs, especially on the calyx (Eastwood 1906;Gonsoulin 1974). Variety californicus occurs in northern California, reportedly extending from Siskiyou Co. southwestward to Lake and Alameda counties and from Yuba Co. south to Fresno Co.; var. fulvescens is mainly southern Californian, reportedly extending from Mendocino and Lake counties, east to Yuba Co., south in the Sierra Nevada to Fresno Co., and San Luis Obispo Co. south to San Diego Co. (Gonsoulin 1974). In contrast, Shevock (1993)  Recent studies of genetic variation within and among populations of Californian Styrax (here designated S. redivivus) using isozymes (Fritsch 1996), and phylogenetic analysis of the genus (Fritsch 1994) have provided insight into the evolutionary history and taxonomy of this group. In this paper, data from these studies are further analyzed and integrated with new data from morphology to gain a deeper understanding of the nature and taxonomic significance of character variation within Californian Styrax. A taxonomic treatment of S. redivivus, a distribution map of historical Californian collections, and a key distinguishing S. redivivus from related taxa are presented. Implications of the data for conservation are discussed.

Distribution Mapping
The specimens cited in Appendix 1 were used to create the distribution map of S. redivivus (Fig. 1). Specimens from the following herbaria were examined: CAS, DS, DUKE, F, GH, JEPS, MO, NY, POM, RSA, SD, and UC. Because populations of S. redivivus are usually local, it was considered useful to arrange specimens by locality, each locality representing a separate population. It was sometimes difficult to determine which collections constituted a population from one locality due to imprecise specimen label information. In such cases, the specimen with the more general information was excluded from the distribution map (Appendix 1). No doubt some independent collections that I have combined into a single locality will, upon further study, constitute separate populations, and vice versa.

Morphology
To assess patterns of morphological variation within S. redivivus, six characters were measured on herbar-..c: -Ol c: ~ angle of base deviation ium specimens: maximum leaf width ( = WIDTH), deviation from longitudinal symmetry (a measure of leaf shape; = SHAPE), hair density on the abaxial lamina surface ( = LF HAIR), arm length of the longest stellate hairs on the abaxial lamina surface ( = ARM-LENGTH), calyx hair thickness ( = CALYX HAIR), and leaf base angle ( = LF BASE). The characters WIDTH, SHAPE, LF HAIR, CALYX HAIR, and LF BASE reportedly distinguish the varieties of S. redivivus. The only other character reported for this purpose, the presence of rufous calyx hairs, was excluded from the analysis after preliminary observations revealed the obvious sporadic nature of this character throughout California. This agrees with Shevock (1993). The character ARMLENGTH, although not reported as a distinguishing character, was included after preliminary observations suggested that hair length might be consistently greater in southern Californian populations than those in northern California. WIDTH equals the distance from the margin to the midrib at the point of maximum lamina width (mm; Fig. 2). SHAPE equals Ucmaxwictthl-l/2)/l, where lcmaxwictth) is the distance from the base of the lamina to the point of maximum width (mm) and l is the maximum length 207 of the lamina (mm; Fig. 2). LF HAIR equals the number of stellate hair bases occurring within approximately 0.14 mm 2 , estimated at 60X. ARMLENGTH (JJ.m) was measured at 60X and excluded hairs of the vein surface, which are sometimes but not always longer than those on the lamina surface. CALYX HAIR was measured by scoring the thickness of the tomentum from 0 to 4, with 0 the thinnest. LF BASE was measured as the angular deviation from a fiat (truncate) leaf base (excluding any attenuate portion often present just above the petiole; Fig. 2).
Leaves on one herbarium specimen from each of 52 of the 83 localities (Appendix 1) were used for measurements. These localities spanned the geographical range of S. redivivus. Measurements were taken on two leaves per individual and averaged, and were standardized by using only material at or just after anthesis and measuring the two largest terminal leaves from fertile shoots on each individual. The use of flowering material facilitated the measurement of calyx pubescence, which often becomes more difficult to assess in the fruiting stage.

Isozymes
Twenty individuals per population from ten populations sampled across the geographical range of S. redivivus were used for isozyme analysis (Appendix 1; Fig. 1 ). For details regarding collection methods, enzymes surveyed, protein extraction, electrophoresis, staining protocols, and interpretation of isozyme banding patterns, see Fritsch (1996).

Data Analysis
Standard genetic identity (I) and distance (D) were calculated from allele frequencies of 24 isozyme loci (Fritsch 1996) for all pairwise population comparisons using the methods of Nei (1972Nei ( , 1978 as implemented with the program GENESTAT Version 2.1 (by P. Lewis and R. Whitkus; Whitkus 1988; Table 1). The distance values were used as input for a permutation test employing Mantel's Z statistic (Mantel 1967) to test concordance between matrices derived from geographic distance and Nei's genetic distance. Matrices compared were 1) genetic distance versus geographic distance and 2) genetic distance versus latitude. Conventional correlation tests are not appropriate in these cases because the entries in each matrix are not independent; the Z statistic tests the strength of association of paired elements in the two matrices by comparison to a null distribution produced by permutation (Dietz 1983). Permutation tests employing Kendall's Kc statistic and Spearman's R statistic were also performed because Z is highly dependent on the specific distance measure used (Dietz 1983). These three statistics together cover a wide range of sampling distributions; ALISO thus, their combined use provided a robust test of matrix association. Four thousand iterations were used for each analysis, and significance was tested at the 0.05 level. All three test statistics were calculated using a computer program by E. J. Dietz (1983). Principal coordinates analysis (PCO), as implemented with NTSYS-pc version 1.8 byE J. Rohlf (Applied Biostatistics Inc., 1993) was used to ordinate genetic differences among populations of S. redivivus. Input consisted of allele frequencies of polymorphic loci. The data were converted to a Nei's genetic distance (Nei 1972) matrix using the SIMGEND command and double-centered using the DCENTER subroutine; eigenvalues were calculated using the EIGEN command; and graphs projecting the populations onto the first two principal coordinates were constructed using the MXPLOT command. To estimate the overall degree of genetic relationship among the populations, the neighbor-joining algorithm (Saitou and Nei 1987) was implemented with the NEIGHBOR81 program in the computer package PHYLIP (PHYLogeny Inference Package by J. Felsenstein, 1989). Input consisted of Nei's genetic distance (Nei 1972; Table 1) for all pairwise comparisons. Analyses were conducted both including and excluding isozyme data from four populations of a species complex from Texas and northeastern Mexico comprising S. platanifolius Engelm. ex Torr., S. texanus Cory, and S. youngiae Cory (the S. platanifolius group [ = the S. texanus group of Fritsch 1996]; complete data set available upon request; for data summary see Fritsch 1996). This group occurs in western Texas and northeastern Mexico and is in all likelihood the closest relative of S. redivivus (Fritsch 1994(Fritsch , 1996. Data from the S. platanifolius group was included to root the neighbor-joining tree. Bootstraps were performed with the programs SEQBOOT, GENDIST, NEIGHBOR, and CONDENSE, as directed in PHYLIP. The morphological characters were tested for associations with latitude using Pearson's correlation test, as implemented by the computer program SYSTAT version 5.2 (SYSTAT, Inc., 1992). A multiple Pearson correlation tested for associations among the characters and latitude. All correlations were tested for significance at the 0.05 level. A principal components analysis (PCA) using all characters except WIDTH was performed to ordinate morphological differences among individuals. The character WIDTH was excluded from all analyses because initial results showed that it was variable throughout the range of S. redivivus (Fig. 5).
On the basis of morphological variation associated with latitude (see Results), the individuals were partitioned into two groups roughly corresponding to northern and southern individuals. These two groups were subjected to further analyses. Pearson correlations of each character versus latitude were performed to compare patterns of variation within the northern and southern groups. The correlation coefficient from the northern group was compared to that of the southern group for each character, as in Zar (1974;tests performed manually). In addition, multiple Pearson correlations were conducted separately for the northern and southern groups.
The individuals were also partitioned into two groups with the Kmeans clustering algorithm (Hartigan 1975) using SYSTAT. This method finds the best way to divide individuals into a specified number of groups so that they are separated as well as possible. Because only two intraspecific taxa have been proposed within S. redivivus, I specified two groups for the algorithm.
To test for group distinctness, canonical discriminant analysis was used to calculate pairwise Mahalanobis distances among both sets of groups (one set from geography, the other from the Kmeans algorithm) using all characters except WIDTH (significance tested at the 0.05 level). In this procedure, the proportion of individuals misclassified by the discriminant function reflects its ability to distinguish among groups.

RESULTS
Permutation tests showed a significant correlation between genetic and geographic distance matrices (Z The PCO analysis of isozyme allele frequencies showed that populations are ordered with respect to latitude along the first principal coordinate, and suggested two clusters (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)Fig. 3). In the neighbor-joining analysis that excluded populations of the S. platanifolius group, the longest branch observed was that connecting the most northerly population samples (Shasta 1, Shasta 2, Butte, and Lake) to the rest; the bootstrap corresponding to this branch was relatively high (76%; Fig. 4). In the neighbor-joining analysis that included populations of the S. platanifolius group, two groups were formed when the tree was rooted with one of the S. platanifolius populations, one comprising populations Shasta 1, Shasta 2, Butte, and Lake, the other comprising Fresno, San Bernardino 1, San Bernardino 2, San Diego 1, San Diego 2, and Santa Barbara (tree not shown).
Five of the six characters (SHAPE, LF HAIR, ARMLENGTH, CALYX HAIR, and LF BASE) showed a significant association with latitude ( Fig.  5B-F; Table 2A). The character WIDTH was variable throughout the range of S. redivivus (Fig. 5A), and therefore was excluded from all subsequent analyses. The five characters were tested for multiple correlation with latitude. When all five characters were included, associations between LF BASE and all other charac- ters except SHAPE were not significant; all other associations were significant except for that between SHAPE and ARMLENGTH, and SHAPE and LF HAIR (Table 2B). When SHAPE was removed from the analysis, all associations were significant except for all associations with LF BASE (not shown); when LF BASE was removed from the analysis, all associations were significant except for those between SHAPE and LF HAIR, and SHAPE and ARM-LENGTH (not shown). When both of these characters were excluded from the analysis, all associations were significant (Table 2C). The characters in Table 2C comprise the only combination of more than two characters and latitude that showed significant associations for all pairwise comparisons.
Examination of the raw data matrix (not shown) showed no clear break in character-state variation for any of the six characters with respect to latitude, and no two characters showed obvious correlated gaps in variation. This is reflected in the PCA with morphological characters, in which distinct clusters of individuals were not evident (Fig. 6). However, PCA was consistent with isozyme data and Pearson correlations in detecting a strong latitudinal trend, with southern individuals tending toward positive scores and northern negative on factor 1.
The distribution of individuals in the scatter plots of three morphological characters versus latitude (SHAPE, LF HAIR, and ARMLENGTH; Fig. 5B-D) suggested a pattern of strong clinal variation among northern individuals, but not among southern individuals. Individuals were partitioned into two groups  on the basis of this distinction and the two groups were analyzed for character correlations separately. None of the Pearson correlations of each character versus latitude in the southern group were sig-  nificant; all correlations in the northern group were significant except for that between WIDTH and LF BASE (Table 3A). Multiple correlation excluding WIDTH and LF BASE showed nonsignificant associations in all pairwise associations between characters for southern individuals (Table 3B). For northern individuals, all pairwise associations were significant except for that between LF HAIR and ARMLENGTH (Table 3C). The correlation coefficients for SHAPE, LF HAIR, ARMLENGTH, and CALYX HAIR were significantly different between the northern and southern groups (Table 3A).
Mahalanobis distances among the southern and northern groups (Fig. 6) as calculated from the canonical discriminant analysis with the five characters were significant (all P [univariate and multivariate] :::; 0.01). The two groups were identified correctly 92% and 86% of the time, respectively, on the basis of the discriminant function (Table 4A). The groups constructed with Kmeans were similar but not identical in composition to those delimited on the basis of geographic location (Fig. 6). The Kmeans algorithm grouped individuals 25, 26, 27, and 37 (see Appendix 1) with the southern individuals, and grouped individual 4 with the northern individuals. The two groups delimited through Kmeans were also significant (all P :::; 0.011). These two groups were identified 96% (group A) and 100% (group B) of the time (Table 4B).

Field Observations
During collecting trips for isozyme analysis made in 1993, I informally assessed the regional abundance of S. redivivus. In most cases, the precise collection locality had to be rediscovered in order to obtain samples-surrounding areas appeared not to harbor the plant, at least in easily accessible places. Regions in which the plant appeared to be so common that precise locality information was not necessary were the Santa Ynez Mountains near San Marcos Pass (Santa Barbara Co.), the Redding area (Shasta Co.), and Palomar Mountain (San Diego Co.). I have been informed (S. Boyd, Rancho Santa Ana Botanic Garden, pers. comm.) that it is also abundant in the Agua Tibia Wilderness, San Diego Co.

Taxonomic Implications of Patterns of Variation within Styrax redivivus
Several patterns elicited by this study provide circumstantial evidence for the delimitation of two infra-211 specific taxa within S. redivivus: 1) correlations and permutation tests indicate the association of several morphological characters and genetic variation with geography and, more specifically, latitude; 2) at least SHAPE and LF HAIR have a genetic basis, as is evident from a common garden study at Rancho Santa Ana Botanic Garden in southern California (pers. obs.); it is likely that all the observed phenotypes associated with latitude are heritable; 3) the neighborjoining trees and the PCO analysis resulting from the isozyme data set suggest the presence of two groups, one northern and one southern; 4) significant differences in the strength of association of four characters (SHAPE, LF HAIR, ARMLENGTH, AND CALYX HAIR) with latitude exist between northern and southern individuals, suggesting that different evolutionary processes have taken place in the two groups; and 5) when two groups are delimited through the Kmeans algorithm, Mahalanobis distances are significant and the discriminant function derived from five morphological characters performs extremely well in predicting group membership.
Nonetheless, there are problems with accepting infraspecific taxa on the basis ofthese criteria. Points (1) and (2) say nothing about gaps or significant differences between groups, which are needed to justify circumscription. Point (3) shows two groups, but little confidence can be afforded to these data for circumscription, because of the low number of population samples.
Point ( 4) clarifies the distribution of variation from north to south. The characters SHAPE, LF HAIR, and ARMLENGTH show strongly clinal variation in the northern individuals, but not in the southern individuals. However, overall association with latitude for these characters is maintained, despite great overlap between southern and northern individuals, because the southern individuals possess the extremes of variation (Fig. 6). The strongly clinal variation seen among the northern populations might be a result of dispersed introgression of genes from southern populations. In this regard, some Californian plant species have expanded their ranges northward from southern refugia in response to a warmer, dryer climate during the Xerothermic (Raven and Axelrod 1978), and this might have happened in southern populations of S. redivivus. Evidence of secondary contact would provide an argument for the delimitation of taxa. However, in the absence of additional molecular variation patterns (i.e., chloroplast and nuclear ribosomal DNA) it remains   Probably the least ambiguous evidence for the circumscription of two taxa comes from the Kmeans algorithm (Point 5), which separates two groups that perform nearly perfectly in the discriminant analysis. Nonetheless, recognition of these groups would require the placement of one specimen from San Diego Co. (Reed 1 0850) in an otherwise northern group, one specimen from Lake Co. (L. Benson 124) in an otherwise southern group, and the ability to distinguish among the taxa in Calaveras Co., where both taxa would occur in close proximity. The two anomalies plus the problem of distinguishing the taxa in clearly intermediate areas does not provide sufficient evidence for recognizing two taxa. The scatter plots of Fig. 5B-E strongly suggest that the significant associations between the characters SHAPE, LF HAIR, ARMLENGTH, and CALYX HAIR result from truly clinal variation with respect to latitude; poor sampling at intermediate latitudes explains the two apparent clusters of data points in several of the plots. Furthermore, examination of the raw data set reveals no instances of any two characters with gaps occurring among the same individuals. Finally, a clear break in character variation is not exhibited in the PCA analysis; rather, the PCA plot shows continuous distribution of individuals ordered with respect to latitude from one extreme of the plot to the other. Fig. 6. Principal components analysis using five morphological characters. The first two axes account for 56.3% and 19.6% of the total variation, respectively. Individuals are numbered from 1 to 52 in order of increasing latitude. Boldface numerals are southern individuals. The line delimits the two groups constructed by the Kmeans algorithm (see text). Table 3. Pearson correlation coefficients (r) and Bonferroni probabilities (P) for morphological characters of Styrax redivivus versus latitude, partitioned into southern (individuals 1-24) and northern (individuals 25-52) groups. For explanation of character abbreviations, see text. Values above the diagonal in multiple correlation tables show P; values below the diagonal show r. (A), single Pearson correlations between each character and latitude; notation at end of row indicates significance level of a test for differences between northern and southern correlation coefficients: *, P < 0.005; **, P < 0.001; N.S., not significant at level 0.05; (B), multiple Pearson correlation between SHAPE, LF HAIR, ARMLENGTH, CALYX HAIR, and LATITUDE, southern individuals; (C), same as (B), northern individuals.  These patterns effectively preclude the delimitation of two taxa in S. redivivus.

Species Status of Styrax redivivus
Evidence from Fritsch (1996) has provided strong justification for recognizing S. redivivus as a distinct species, regardless of whether an isolation (biological; Mayr 1969), phylogenetic (Rosen 1978;Mishler and Donoghue 1982), or taxonomic (Cronquist 1988) species concept is employed. Genetic distances derived from isozyme variation indicate high genetic divergence between Californian and Mediterranean populations (Fritsch 1996). This provides inferential evidence for species status according to the isolation species concept. Divergence time estimates based on genetic distances indicate that S. redivivus and S. officinalis have probably been separated intercontinentally for at least five million years (Fritsch 1996).
Phylogenetic evidence indicates that the two entities are not sister taxa: the closest relatives of S. redivivus   are the species comprising the S. platanifolius group, as inferred from studies of chloroplast and nuclear ribosomal DNA (Fritsch 1994(Fritsch , 1996. The current treatment (i.e., including S. redivivus inS. officinalis while retaining the Texas-Mexico taxa as different species) is unambiguously inconsistent with the phylogeny. To be consistent with the phylogeny it is justifiable to either 1) include S. youngiae, S. texanus, S. platanifolius, and S. redivivus in S. officinalis, or 2) remove S. redivivus from S. officinalis. The first circumscription might be justified because Mediterranean Styrax is the sister group to the California-Texas-Mexico clade and thus all these taxa (the S. officinalis group) comprise a monophyletic group (Fritsch 1994(Fritsch , 1996. However, isozyme genetic divergence between California and Texas-Mexico populations is within the range typical for that between congeneric species rather than within the same species (Fritsch 1996). Final evidence comes from the elucidation of diagnostic morphological characters from field observations and examination of herbarium specimens, providing justification for the species status of S. redivivus under a taxonomic species concept. My field observations show that S. redivivus has orange-yellow pollen whereas S. officinalis has light yellow pollen. Also, S. redivivus has corolla lobes that are reflexed an average of 35° away from the longitudinal plane at anthesis, whereas S. officinalis has corolla lobes that are refiexed an average of 80° (N = 40 for both species, two populations and 20 plants per population sampled for each species, P < 0.001). These two diagnostic characters were only noticed by comparing both groups in the field, which probably explains why they were not detected by previous workers.
Examination of herbarium specimens of S. officinalis (F, GH, MO, NY, RSA) provides additional mor-phological evidence for treating S. redivivus as a distinct species. Torrey (1853) reported three characters distinguishing S. redivivus from S. officina/is: thickened pedicels, a longer staminal tube, and fewer-flowered racemes. The first two characters are clearly diagnostic (see Appendix 2); the third is not diagnostic because of overlap, but on the average the species are different ( 1-6 flowers in S. redivivus, mean = 2.8, 2-6 flowers in S. officina/is, mean = 3.8). Munz and Johnston (1924) listed three characters distinguishing S. redivivus from S. officina/is: dark-versus light-colored stalked stellate hairs on the leaf veins and midrib abaxially, a more compressed style, and a more prominently lobed stigma. Apparently, they either did not see Torrey's (1853) description or they ignored it. All these characters they considered to be minor and inconstant; hence, they treated S. redivivus as a variety of S. officina/is. Of these three characters, the hair color difference is diagnostic (see Appendix 2); the other two are not at all reliable. The only other character reported to distinguish the two taxa is larger flowers in S. redivivus (Rehder 1915;a varietal distinction). This character also cannot be considered diagnostic, although the flowers are generally larger (16-26 mm long in S. redivivus, 13-20 mm long in S. officina/is). I have discovered two other characters useful in distinguishing the taxa: pedicel length ( 4-9 mm long in S. redivivus, 8-17 mm long in S. officina/is, and relative pedicel versus calyx length (pedicel 0.5-1.4 times as long as the calyx in S. redivivus, 1.4-3.3 times in S. officina/is). These taxa can easily be distinguished from each other and from the taxa of the S. platanifolius group with the aid of the key in Appendix 2.

Conservation
Styrax redivivus apparently has a relict distribution: its ancestors were part of a more widely distributed species in the Tertiary (Fritsch 1996). Yet, the current ecological factors that limit the distribution and abundance of this species are poorly known. The species appears to tolerate widely different soil types, shade levels, and moisture regimes, although it seems to prefer clay soils or rocky outcrops, protected ravines or washes, and north-or east-facing slopes. Nevertheless, in many areas seemingly suitable for the plant, it is not present. It appears not to be establishing new populations in any significant number; some populations are senescent (e.g., Auburn, Placer Co.). The only locality where I observed seedlings was Mt. Palomar (San Diego County); they were growing underneath presumably maternal parents. Plants probably resprout after fire; they do not spread vegetatively. Dispersal mechanisms are unknown, and seeds have no structures associated with dispersal via wind or water; therefore, seeds may be dispersed only locally by ei-ALISO ther gravity or ground mammals, or both; if so, the seeds are probably only secondarily preferred (as suggested by one common name for the plant, "bitternut" (Jepson 1925). Seeds may also be water-dispersed downslope. Detailed study of the natural history and ecology (especially seed ecology, demography, and habitat requirements) of S. redivivus are needed to provide further insight into the reasons for the species' sporadic distribution, Californian endemism, and role in Madrean-Tethyan vegetation patterns (see Fritsch 1996), and would also facilitate conservation efforts.
Because of its local distribution and apparently poor establishment capacity, S. redivivus should be monitored for possible population decline. The species is still common enough to forego state or federal listing, but many, especially low-elevation, populations are threatened by overgrazing, dam building, and urbanization. I have not directly observed browsing by grazing animals, but I have seen the heavily grazed slopes at Kings River (Pine Flat Reservoir, Fresno Co.) that were nearly devoid of the plants except for rocky areas inaccessible to livestock. The related species S. texanus has declined dramatically through the effects of exotic grazing mammals in Texas (Cory 1943;Cox 1987) and S. officina/is serves as forage for grazing mammals in the Mediterranean region (Le Houerou 1981; pers. obs.). Because S. redivivus prefers water courses along foothill woodland slopes, reservoir creation has played a role in its decline. The establishment of Pine Flat Reservoir on the Kings River is known to have reduced the abundance of S. redivivus in the area, as is evident by collections from now inundated areas. At Folsom (Sacramento Co.) and Silver Rapids (Calaveras Co.), inundation has no doubt contributed to local decline or extirpation. Urbanization also appears to have contributed to the probable elimination of the Folsom population, as well as populations near Clear Lake, Lake Co. Styrax redivivus must have at one time been abundant at Clear Lake, judging from the numerous collections made, but now appears to be restricted; I only found one small population after several hours of searching along the western lake perimeter.
It is hoped that this report will increase awareness of the uncommon nature of S. redivivus enough so that vouchers will be made when the plant is encountered, especially at localities not listed here or at localities where the species is thought to be extirpated. Wilbur for comments and suggestions on the manuscript; and J. T. Columbus and an anonymous reviewer for critical reviews. Also thanks to M. Debacon for laboratory assistance. This work was supported in part by NSF Doctoral Dissertation Improvement Grant DEB-9310979, a Hardman Foundation, Inc. grant, the Rancho Santa Ana Botanic Garden, and the A. W. Mellon Foundation.