Bot. Bull. Acad. Sin. (2002) 43: 131-138

Whang et al. — A morphometric analysis within the Ixeris chinensis complex

A morphometric analysis of infraspecific taxa within the Ixeris chinensis complex (Asteraceae, Lactuceae)

Sung Soo Whang¹, Kyung Choi^2,5, Robert S. Hill³, and JaeHong Pak^4,*

¹Division of Science Education, Chonbuk National University, Chonju 561-756, Republic of Korea

²Department of Biology, Yeungnam University, Keongbuk 712-749, Republic of Korea

³Department of Environmental Biology, University of Adelaide, SA 5005, Australia

⁴Department of Biology, Kyungpook National University, Taegu 702-701, Republic of Korea

(Received July 17, 2001; Accepted October 3, 2001)

Abstract. Morphometric methods were used with 40 morphological characters from each of 110 individuals collected from 11 populations of the Ixeris chinensis complex to test hypotheses regarding infraspecific taxa. Individuals of the nearest relative I. tamagawaensis were included as an outgroup. To utilize structual associations of the original data and the cumulative values of character variation, a modified data set was generated by matching factor scores with the large original data set. As a result, the original data was reduced to about 10% of its original size. The modified data also emphasized the cumulative values of character variations. Structural associations among characters measured, however, were not significant for most factors extracted even though some obvious trends appeared. Relationships between cumlative values of variation and taxa in the complex are discussed based on the results of the canonical discriminant function analysis, ANOVA, multiple comparison test, and cluster analysis, using the modified data. This study supports Kitamura's (1956) classification of infraspecific taxa in the I. chinensis complex. Given that there is no overlap of cumulative values of variation between the species complex and I. tamagawaensis, there is probably a 75% character overlap within the complex.

Keywords: Ixeris chinensis complex; Infraspecific taxa; Morphometric analysis.

Introduction

The term "species complex" has frequently been applied to flowering plants, but its origin is obscure. It has often been used to describe species aggregations made up of infraspecific taxa such as microspecies, agamospecies, and semispecies (e.g. Fosberg, 1942; Sylvester-Bradley, 1952; Davis and Heywood, 1963; Northington, 1976; Grant, 1981). Several criteria have been used to establish groups of infraspecific taxa, including their morphological discreteness, ecological differences, genetic divergence, and natural reproductive isolation (e.g. Davis and Heywood, 1963; Gilmartin, 1974; Grant 1981; Stuessy, 1990). Morphological discreteness plays a major role in the early stage recognition of infraspecific taxa. A practical difficulty here is that almost all morphological characters and/or character states used for evaluating infraspecific taxa are only slightly differentiated from one another and usually show considerable overlap. Therefore the identification of infraspecific taxa and their hierarchical arrangement within a complex of species usually requires a combination of

several morphological characters, especially if other forms of data are not available.

The Ixeris chinensis complex (Asteraceae, Lactuceae) is distributed from Siberia to Japan, including Korea, Taiwan, and China (Pak et al., 1997). This complex has a basic chromosome number of eight (Pak and Kawano, 1990), but also contains diploids (2n = 16), triploids (2n = 24), and tetraploids (2n = 32) (Ishikawa, 1921; Babcock et al., 1937; Chuang et al., 1962; Hsu, 1967, 1970; Peng and Hsu, 1978; Pak and Kawano, 1990) and exhibits sympatry and/or allopatry among the ploidy (e.g. Pak et al., 1995, 1997). The diploid plants observed are sexual while triploids and tetraploids are asexual. Many complicated variations in external morphology have been reported in hybrids between sexual and usually asexual plants, which can revert to sexual reproduction under some conditions (e.g. Harlan and de Wet, 1963; Yahara, 1983, 1990). Within this complex a complicated morphological overlap, without any discontinuities, has led to taxonomic difficulty (Kitamura, 1956; Pak and Kawano, 1992; Kim et al., 1999). Kitamura (1956) classified this complex into the three subspecies Ixeris chinensis (Thunb.) Nakai subsp. chinensis, subsp. strigosa (H. Lév. et Vaniot) Kitamura, and subsp. versicolor (Fischer) Kitamura. Tzvelev (1964) classified it as three different species in a different genus, Ixeridium chinensis (Thunb.) Tzvelev, I. strigosum (H. Lév. et Vaniot) Tzvelev and I. gramineum (Fisch.) Tzvelev. Pak and

⁵Current address. Division of Life Science & Graduate School of Biotechnology, Korea University, Seoul 136-701, Republic of Korea.

*Corresponding author. Tel: +82-53-950-5352; Fax: +82-53-953-3066; E-mail: jhpak@kyungpook.ac.kr

Kawano (1992) classified it as two subspecies of one species and one different species in the same genus; Ixeris chinensis subsp. chinensis and subsp. versicolor and I. strigosa (H. Lév. et Vaniot) Pak et Kawawo. The taxonomic resolution of this complex, therefore, remains uncertain.

In an attempt to resolve difficulties caused by morphological overlap, multivariate methods have increasingly been applied, especially in species complexes of flowering plants (e.g. Sokal and Sneath, 1963; Gilmartin, 1967; Jensen and Eshbaugh, 1976; McNeil, 1984; Jensen et al., 1993). Several multivariate analyses are applicable to this problem, and the adoption of different statistical procedures in dealing with such problems is common. The classificatory schemes proposed for the Ixeris chinensis complex based on morphological characters can be tested using multivariate statistical analysis of the huge amount of raw data that can be collected from such a morphologically variable group. As a first step in such a process, a factor analysis capable of regenerating reduced data might be effective in order to yield a clear output in later treatments.

This study aims to establish a theoretical model for the categories of infraspecific taxa and to provide taxonomic resolution of the Ixeris chinensis complex based on morphometric analyses. We will not only discuss categories at the level of both species and infraspecies, but we will also test the taxonomic schemes previously proposed for this complex.

Materials and Methods

Specimens of the Ixeris chinensis complex were selected from among 81 populations, chosen on the basis of ploidy and geographic distribution (Pak et al., 1997, Figure 1). Twenty individuals from two populations of I. Tamagawaensis (Makino) Kitamura were collected near Tamagawa, Ozaku, Oome-shi, Tokyo, Japan and were added to the analysis because this species is regarded as the nearest relative of I. chinensis (Kitamura, 1956; Pak and Kawano, 1990, 1992, Figure 1). The complex has generally been recognized as comprising three taxa although their taxonomic category and nomenclature differs among researchers (Kim et al., 1999). To begin with, Kitamura's (1956) two subspecies of I. chinensis were recognized, with subspecies chinensis only being diploid and subspecies strigosa being both triploid and tetraploid. Individuals of the triploid and tetraploid populations of I. chinensis subsp. strigosa were identified and separated for morphological analysis. Kitamura's (1956) Ixeris chinensis subsp. versicolar, distinguished within the complex by having a different corolla color and branched roots, is restricted to North Korea, Manchuria, and Russia, and was omitted here due to lack of sufficient material for comparison. Three representative populations of each ploidy out of the 81 populations documented (Pak et al., 1997) were selected, and ten specimens from each population were investigated. The specimens were obtained from field col

Figure 1. Map showing collecting localities of the Ixeris chinensis complex and its nearest relative I. tamagawaensis. ~=I. chinensis subsp. chinensis (Mt. Keumo, Janghang-ri and Mokye-ri in Kyungbuk); p=I. chinensis subsp. strigosa (triploid) (Moraejae and Malgoljae in Chonbuk, Damyang-up in Chonnam); ¢=I. chinensis subsp. strigosa (tetraploid) (Is. Nami and Taebong-ri in Kyunggi, Sanchong-up in Kyungnam), ê=I. tamagawaensis (near Tamagawa in Tokyo).

lections and from the following herbaria: KYO, SKK, SNAU, TI, TNS, KNU, Chonnam National University, and Kangwon National University (Figure 1).

Forty characters were scored for the morphometric analysis (Table 1). They were measured using a digimatic caliper and hand lens. Of these characters, radical leaf blade length (RLL) was measured from three leaves in order of size among those attached at the base; internode length (INL) was measured at the second node up from the base; length and width of inner-phyllary (IIL and IIW) were measured on three randomly selected leaves; length and width of the outer-phyllary (OIL and OIW) were measured on three replicates; and length and width of the achene (ACL and ACW) were measured on one mature specimen.

Factor analysis was used to interpret structural associations among the 40 morphological characters, as well as to reduce the large amount of data measured. Factors >1 in eigenvalue were selected and scored. An analysis of variance of the modified data, generated by matching the factor score (åfactor score coefficients of character measurement to factor 1 × standard scores of taxa to characters) of the original data, was performed to determine whether significant differences among members of the I. chinensis complex and its nearest relative I. tamagawaensis exist or not. Multiple comparison by four different methods, Duncan, Tukey-HSD, Tukey-B and Scheffe test, was conducted on the modified data for the first two factors. Canonical discriminant functions were also conducted to ordinate relationships between variables

and cumulative values of variation produced by factor scores. A cluster analysis using average linkage was performed on the modified data to evaluate similarity among variables and to identify taxa. All significance levels were determined at p £ 0.05. The numerical analysis was performed using SPSS for MS WINDOWS, Version 6.1.

Results

For the morphometric analysis of the Ixeris chinensis complex, 40 characters of external morphology were measured from each of 110 individuals collected from 11 populations. After a factor analysis, the 10 factors that had eigenvalues that summed to > 1 were extracted (Table 2). Total cumulation of the extracted factors accommodated 78.1% of variation in the original data. These factors were characterized with respect to morphological characters that had high absolute values for factor loading. Factor loadings indicate the overall importance of each character, while eigenvalues for each factor indicate the portion of total variance accounted for by that factor. In distinguishing high factor loading, loading with values ³ 0.5 of the absolute value were arbitrarily adopted as the biological standard.

The fifth to tenth factors contained only one or no characters with a factor loading > 0.5 in absolute value. Furthermore, these factors had eigenvalues < 2, and accounted for only 16.4% of the original data (Table 2). The first four factors, however, contained several morphological characters with factor loadings > 0.5 in absolute value. Each of these factors also had high eigen values (> 3), and they accounted for about 60% of the original data (Table 2). Accordingly, the focus is mostly on aspects of the first four factors.

Factor 1 had high loadings for 14 characters; first, second and third radical leaf blade length in the tooth region (FRLWT, SRLWT, TRLWT), numbers of cauline leaves, internodes and outerphyllaries (NCL, INL, NII), cauline leaf blade width in the tooth region (CLWT), first, second and third inner-phyllaries length (FIIL, SIIL, TIIL), involucre and achene length (IVL, ACL), and length and width of floret (FLL, FLW). The structural relationships among these characters were complex and scattered, but the factor was strongly influenced by several related groups of characters such as the length of radical leaves in the tooth region, the length of inner-phyllaries, and the length and width of florets (Table 2). Factor 2 had high loadings for 10 characters; plant height (PLH), first, second and third radical leaf blade length (FRLL, SRLL, TRLL), number of capitula per plant (NFPI), first, second and third outer-phyllaries length (FOIL, SOIL, TOIL), second outerphyllary width (SOIW), and number of pappus setae per floret (NPPF). This factor was also difficult to circumscribe because of complex structural relationships among characters. However, two groups of characters, the lengths of radical leaves and outerphyllaries, were characteristic of this factor (Table 2). Factor 3 had high loadings for two characters, the first and third outer-phyllaries

Figure 2. Canonical discriminant functions using the four factor scores of the Ixeris chinensis complex and its nearest relative I. tamagawaensis (Group 2: I, chinensis subsp. chinensis; Group 3: I. chinensis subsp. strigosa (triploid); Group 4: I. chinenisis subsp. strigosa (tetraploid); Group 5: I. tamagawaensis).

Botanical Bulletin of Academia Sinica, Vol. 43, 2002

Whang et al. — A morphometric analysis within the Ixeris chinensis complex

width (FOIW, TOIW). These characters, therefore, were correlated with the width of outer-phyllaries, even though confidence in this relationship is lower because the second outer-phyllary was not included (Table 2). Factor 4 had high loadings for five characters; first, second and third radical leaf blade width in the tooth free region (FRLWF, SRLWF, TRLWF), cauline leaf width in the tooth free region (CLWF), and involucre width (IVW). These characters came from three different parts of the plant, and were apparently related to leaf width development. The character most strongly influencing this factor was the width of radical leaves in the tooth free region (Table 2).

To examine the relationship between variables and cumulative values of variation generated, a canonical discriminant function was ordinated with modified (compressed) data (Figure 2). All data for Ixeris tamagawaensis were clustered in a single group and clearly isolated by function 1 from the values of other variables. However, some data for I. chinensis subsp. chinensis and I. chinensis subsp. strigosa were not isolated from each other but overlapped to some extent in both function 1 and 2. For functions 1 and 2, triploid and tetraploid populations of I. chinensis subsp. strigosa overlapped almost completely (Figure 2).

In order to estimate whether there are significant differences within variables, an analysis of variance was performed on the modified data (Table 3). The taxa differed significantly in the first three factors, those with the highest eigenvalues and covering about 50% of the initial data. However, they did not differ significantly in factors 4, 6, 7, 9 and 10, which appears to reduce the implications that can be drawn from the original data (Table 3).

The analysis of variance demonstrates that the variables are quite different in their significance levels. To determine these levels, various multiple comparison tests were conducted on the modified data for factors 1 and 2 (Table 4). Ixeris tamagawaensis, which was included as an outgroup, differs significantly from all other taxa in the I. chinensis complex. Furthermore, within the complex I. chinensis subsp. chinensis was significantly different from both triploid and tetraploid populations of I. chinensis subsp.

Figure 3. Average linkage dendrogram for the members of the Ixeris chinensis complex and its nearest relative I. tamagawaensis (Ch: I. chinensis subsp. chinensis; S3: I. chinensis subsp. strigosa of triploids; S4: I. chinensis subsp. strigosa of tetraploids; Ta: I. tamagawaensis).

strigosa. However, triploid and tetraploid populations of I. chinensis subsp. strigosa were not significantly different from one another (Table 4).

To investigate similarity among the variables considered, a cluster analysis was performed on the modified data (Figure 3). Triploid and tetraploid populations of I. chinensis subsp. strigosa clustered first at a rescaled distance of < 2. Ixeris chinensis subsp. chinensis formed the second cluster with I. chinensis subsp. strigosa at a rescaled distance of < 7. Lastly, the outgroup, I. tamagawaensis, clustered at a rescaled distance of about 25 (Figure 3).

Discussion

The Ixeris chinensis complex is exceedingly polymorphic, and this is reflected in its diverse taxonomic treatment (e.g. Kitamura, 1956; Tzvelev, 1964; Pak and Kawano, 1992). The complex, however, has generally been recognized as having three taxa even though their taxonomic level and nomenclature has differed considerably among researchers (Kim et al., 1999). Kitamura (1956) described the root morphology, plant height, number of cauline leaves, and involucre and achene length as diagnostic characters of this complex. He circumscribed Ixeris chinensis subsp. chinensis by a plant height of 20-35 cm, 2-4 cauline leaves, involucres 6-8 mm long, achenes 4-6 mm long, yellow corolla, and roots that branch often. Ixeris chinensis subsp. strigosa is 25-50 cm in height and has 1-2 cauline leaves, involucres 9-10 mm long, achenes 5.5-7 mm long, whitish ligules with slightly purple surfaces, and roots that are often unbranched. Ixeris chinensis subsp. versicolor is the smallest taxon in the complex at 10-20 cm in height, and has 1-2 cauline leaves, involucres 8-9 mm long, achenes 4.5-5.5 mm long, variously colored corollas, and roots that branch often. These characters described by Kitamura (1956) reveal some degree of overlap among the three taxa of the complex except for corolla color. Tzvelev (1964) revised Kitamura's classification, producing three Ixeridium species, I. chinensis, I. strigosum and I. gramineum. Pak and Kawano (1992) concluded that Ixeris strigosa clearly differs from I. chinensis in having erect stems, a few cauline leaves, and whitish ligules with slightly purple outer surfaces. They revised this complex

Botanical Bulletin of Academia Sinica, Vol. 43, 2002

into two species, Ixeris chinensis and I. strigosa, and further split the former into two subspecies, subsp. chinensis and subsp. versicolor. They also suggested that detailed studies were required to understand the total range of morphological variation in order to provide a more refined taxonomic scheme (Pak et al., 1995, 1997).

In general, if speciation is complete above the population level, new taxa should be compartmentalized not only by discontinuity in morphological characters but also by specific molecular characters. However, if speciation is in process, most morphological characters may overlap rather than exhibit discontinuity, and molecular characters may not be unique. In that case, taxonomic perspective usually relies on a subjective assessment. The term species complex was used to describe species aggregations sharing the specific morphological and molecular features (e.g. Davis and Heywood, 1963; Grant, 1981; Stuessy, 1990; Judd et al., 1999) in taxa that are undergoing speciation. Although their taxonomic affinity may be difficult to determine, some form of taxonomic resolution is desirable.

The aims of this study were to establish a model for infraspecific taxa within the Ixeris chinensis complex based on morphometric approaches and to use this as a test for published taxonomic schemes. As mentioned above, this complex shows considerable overlap in morphological characters (Kitamura, 1956; Pak and Kawano, 1992), and its taxonomic treatment differs among researchers (Pak et al., 1995, 1997). An important point often overlooked by taxonomists is that the cumulative values of character variation occurring among taxa might be taxonomically useful. Although all morphological characters considered overlap considerably, those cumulative values modified from original measurements may be particularly powerful in recognizing difficult taxa when the degree of character variation differs signficantly among taxa (Whang et al., 1998). That point was the focus of this study, which considered both the structual associations of the original data and the cumulative values of variation as morphometric data. The original data set was obtained by the measurement of 40 morphological characters from each of 110 individuals of the I. chinensis complex and I. tamagawaensis. Firstly,

the 10 factors that had eigenvalues that summed to > 1 were extracted with a factor analysis (Table 2). However, structural relationships between characters were indistinct in most factors extracted even though there were some noticeable tendencies. A modified data set was generated by matching factor scores, calculated by the correlation coefficients of each factor, with the large original data set. The initial data set was reduced to about 10% of its original size and was significant both biologically and statistically. The nature of the modified data also emphasized the cumulative values of character variations. The modified data were examined by a series of multivariate methods in order to recognize taxa in the complex.

A canonical discriminant function analysis was conducted to understand the relationships between the four taxa examined (Figure 2). All the cumulative values of I. tamagawaensis were well separated from all other taxa by function 1 (Figure 2). It is likely that, in spite of being the nearest relative, this species had clear morphological discontinuities from the other taxa. However, the cumulative characters did not separate the other taxa with functions 1 and 2, as they overlapped considerably with each other (Figure 2), confirming their close relationship. Tzvelev's (1964) proposal of three species in this complex (which he placed in the genus Ixeridium) is not supported by this analysis. Also, the separation of the complex into two species, I. chinensis and I. strigosa by Pak and Kawano (1992), is not supported here because of the considerable morphological overlap recorded. The results of the ANOVA indicated significant differences for factors 1, 2, 3 and 5, which cover about 60% of the original data (Table 3). Furthermore, several multiple comparison tests, carried out to determine which taxa differed significantly (Table 4), showed that I. tamagawaensis (included as an outgroup) differs significantly from all other taxa, and I. chinensis subsp. chinensis differs significantly from both triploid and tetraploid populations of I. chinensis subsp. strigosa as well as from I. tamagawaensis (Table 4). Therefore, the cumulative character segregation of this complex did not distinguish separate taxa through the canonical discriminant function test, but two taxa were evi

Whang et al. — A morphometric analysis within the Ixeris chinensis complex

dent in the complex using the multiple comparison test. These findings are compatible with the taxonomic classification of Kitamura (1956), who separated I. chinensis subsp. strigosa from I. chinensis subsp. chinensis based on geographic distribution. The cluster analysis, conducted to determine similarities among taxa (Figure 3), shows no cumulative character overlap between the species complex and I. tamagawaensis, but indicates a 75% overlap between I. chinensis subsp. chinensis and I. chinensis subsp. strigosa. There appears to be about a 95% cumulative character overlap between two groups (triploids and tetraploids) of I. chinensis subsp. strigosa.

In conclusion, this study provides morphological support for Kitamura's (1956) classification of infraspecific taxa in the I. chinensis complex. Given that there is no overlap of cumulative values of variation between the Ixeris chinensis species complex and I. tamagawaensis, it is probable that there was 75% overlap within the complex.

Acknowledgements. We thank the directors and curators of KYO, SKK, SNAU, TI and TNS as well as the herbaria of Chonnam National University, Kangwon National University, and Kyungpook National University (recently listed as KNU). This study was partly supported by the academic research fund of the Ministry of Education, Republic of Korea (BSRI '984404).

Literature Cited

Babcock, E.B., G.L. Stebbins Jr, and J.A. Jenkins. 1937. Chromosomes and Phylogeny in some Genera of Crepidinae. Cytologia, Fujii Jub. Vol., pp. 188-210.

Chuang, T.I., C.Y. Chao, W.L. Hu, and S.C. Kwan. 1962. Chromosome numbers of the vascular plants of Taiwan 1. Taiwania 8: 51-66.

Davis, P.H. and V.H. Heywood. 1963. Principles of Angiosperm Taxonomy. Princeton NJ, Van Nostrand.

Fosberg, F.R. 1942. Subspecies and variety. Rhodora 44: 153-157.

Gilmartin, A.J. 1967. Numerical Taxonomy - an eclectic viewpoint. Taxon 16: 8-12.

Gilmartin, A.J. 1974. Variation within populations and classification. Taxon 23: 523-536.

Grant, V. 1981. Plant Speciation. 2nd edn. Columbia University Press, New York.

Harlan, J.R. and J.M.J. de Wet. 1963. Role of apomixis in the evolution of the BothriochloaDichanthium complex. Crop Sci. 3: 314-316.

Hsu, C.C. 1967. Preliminary chromosome studies on the vascular plants of Taiwan (I). Taiwania 13: 117-129.

Hsu, C.C. 1970. Preliminary chromosome studies on the vascular plants of Taiwan (III). The family Compositae. Taiwania 15: 17-29.

Ishikawa, M. 1921. On the chromosome of Lactuca (a preliminary note). Bot. Mag. Tokyo 35: 153-158.

Jensen, R.J. and W.H. Eshbaugh. 1976. Numerical taxonomic studies of hybridization in Quercus. II. Populations with wide areal distribution and high taxonomic diversity. Syst.

Bot. 1: 11-19.

Jensen, R.J., S.C. Hokanson, J.G. Isebrands, and J.F. Hancock. 1993. Morphometric variation in oaks of the Apostle Islands in Wiscons: evidence of hybridization between Quercus rubra and Q. ellipsoldalis (Fagaceae). Amer. J. Bot. 80: 1358-1366.

Judd, W.S., C.S. Campbell, E.A. Kellogg, and P.F. Stevens. 1999. Plant Systematics. A Phylogenetic Approach. Sinauer Associates, Massachusetts.

Kim, N.C., K. Choi, I. Kim, B.B. Seo, S.D. Song, and J.H. Pak. 1999. External morphological analysis of diploids, triploids, and tetraploids of the Ixeris chinensis complex (Asteraceae: Lactuceae). Kor. J. Plant Taxon. 29: 19-35. (in Korean)

Kitamura, S. 1956. Compositae Japonicae, Pars Quinta. Mem. Coll Sci. Univ. Kyoto, Ser. B (Biol.) 23: 116-123.

McNeil, J. 1984. Numerical Taxonomy and Biosystematics. In W.F. Grant (ed.), Plant Biosystematics. Academic Press, Toronto.

Northington, D.K. 1976. Evidence bearing on the origin of infraspecific disjunction in Sophora gypsophila (Fabaceae). Plant Syst. Evol. 125: 233-244.

Pak, J.-H. and S. Kawano. 1990. Biosystematic studies on the genus Ixeris (CompositaeLactuceae). II. Karyological analysis. Cytologia 55: 553-570.

Pak, J.-H. and S. Kawano. 1992. Biosystematic studies on the genus Ixeris (CompositaeLactuceae). IV. Taxonomic treatments and nomenclature. Mem. Coll. Sci. Kyoto Univ., Ser. B (Biol.) 15: 29-60.

Pak, J.-H., N.C. Kim, K. Choi, I. Kim, B.B. Seo, and S.D. Song. 1997. Geographical distributions of diploids and polyploids of the Ixeris chinensis complex (Asteraceae: Lactuceae) in East Asia. Acta Phytotax. Geobot. 48: 187-196.

Pak, J.-H., N.C. Kim, K. Choi, I. Kim, B.B. Seo, and S.D. Song. 1995. Geographical distributions of diploids, triploids, and tetraploids of the Ixeris chinensis complex (Asteraceae: Lactuceae) of South Korea. Kor. J. Plant Taxon. 25: 221-236. (in Korean)

Peng, C.I. and C.G. Hsu. 1978. Chromosome numbers in Taiwan Compositae. Bot. Bull. Acad. Sin. 19: 53-66.

Sokal, R.R. and P.H.A. Sneath. 1963. Principles of Numerical Taxonomy. Freeman, San Francisco.

Stuessy, T.F. 1990. Plant Taxonomy. Columbia University Press, New York.

Sylvester-Bradley, C. 1952. The Classification and Coordination of Infra-Specific Categories. Systematics Association, Mimeograph, London, p. 19.

Tzvelev, N.N. 1964. Ixeridium. In E.K. Vobroop and N.N. Tzvelev (eds.), Flora of the USSR. Vol. 29, Leningrad, pp. 388-393.

Whang, S.S., K. Kim, and W.H. Hess. 1998. Variation of silica bodies in leaf epidermal long cells within and among seventeen species of Oryza (Poaceae). Amer. J. Bot. 85: 461-466.

Yahara, T. 1983. A biosystematic study on the local populations of some species of the genus Boehmeria with special reference to apomixis. J. Facul. Sci., Univ. Tokyo, Sect. 3, Bot. 13: 217-226.

Yahara, T. 1990. Evolution of agamospermous races in Boehmeria and Eupatorium. Plant Species Biol. 5: 183-196.

Botanical Bulletin of Academia Sinica, Vol. 43, 2002