Variation patterns in the Phlomis x composita (Lamiaceae) hybrid complex in the Iberian Peninsula (original) (raw)

Variation patterns in the Phlomis ×\times composita (Lamiaceae) hybrid complex in the Iberian Peninsula

RAFAEL G. ALBALADEJO*, A. APARICIO and S. SILVESTRE
Departamento de Biología Vegetal y Ecología, Facultad de Farmacia, Universidad de Sevilla, 41012 Sevilla, Spain

Received May 2003; accepted for publication September 2003

Abstract

Hybridization seems to be common in the genus Phlomis (Lamiaceae) in the Iberian Peninsula, especially in the P.×P . \times composita complex. In order to detect patterns of morphological variation linked to eco- and geographical variation, ecological (cluster and canonical correspondence) and morphometric (principal component and discriminant function) analyses were performed. Character count procedure was applied to discern between divergence and reticulate events for the origin of the morphologically intermediate plants. Following these analyses clear patterns were detected suggesting the existence of four independent morphological groups also supported by the ecogeographical data. These are P. crinita ssp. crinita (Levante, eastern Spain), P. crinita ssp. malacitana (Andalusia, southern Spain) and P. lychnitis (widespread) being the extremes of the morphological variation, and P.×P . \times composita including all the morphologically intermediate individual plants. Furthermore, at the population level significant differences in hybrid plant frequencies between areas were also found. © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 145, 97-108.

ADDITIONAL KEYWORDS: character count procedure - ecological analyses - hybrid zone - morphometric analyses - natural hybridization - Spain.

INTRODUCTION

Morphological intermediacy in nature can be a consequence of either divergence or reticulate events, and several methods have been developed and used to discern between them (e.g. Thorpe, 1984; Wilson, 1992; Rieseberg & Morefield, 1995). One of the most widely employed approaches using morphological data is the ‘character count procedure’ proposed by Wilson (1992) (e.g. Thébaud & Abbott, 1995; Miller & Spooner, 1996; Hawkins et al., 1999; Levin, 1999). However, Rieseberg & Ellstrand (1993) and Rieseberg (1995) have demonstrated that hybridization does not always result in morphological intermediacy, especially when: (1) introgression is extensive (2) the hybridization process happened a long time ago, and (3) the expression of some morphological characters is not under an additive genetic control (e.g. heterosis or dominance) (Levin, 1999). In all these cases, the application of Wilson’s method would produce a type I error (i.e. reject-

[1]ing the hybridization hypothesis even though it was correct). Therefore, and according to Levin (1999: 270): ‘this is a conservative test: when a hypothesis of hybridization is supported, it is very likely to be correct.’

In addition, multivariate ordination techniques such as principal components analysis (PCA), principal coordinates analysis (PCoA) or discriminant function analysis (DFA) have been widely used in hybridization studies to detect putative parental and hybrid taxa (e.g. Thébaud & Abbott, 1995; Gugerli, 1997; Levin, 1999) and backcrossed individuals in hybrid swarms (Wagner, 1983; Wilson, 1992), despite the fact that they cannot distinguish between hybridization and divergence (Wilson, 1992).

The genus Phlomis L. (Lamiaceae) comprises more than 100 species, primarily distributed in Eurasia and north-western Africa with two main centres of diversity in southern and eastern Anatolia and north-western Iran (Azizian & Moore, 1982). Although only a few taxa of the genus occur in the Iberian Peninsula, they are surrounded by considerable taxonomic, nomen-

1. *Corresponding author. E-mail: albaladejo@us.es ↩︎

clatural and ecogeographical confusion, supposedly due to natural hybridization and the existence of frequent intermediate plants. Phlomis lychnitis L. is widespread across the Peninsula, whereas PP. crinita Cav. is restricted to some limestone areas in southern and eastern Spain (Andalusia and Levante, respectively). Here the two taxa are sympatric and intermediate plants have been recognized morphologically since the early 20th century (Pau, 1918a).

Pau (1925) recognized both P.lychnitis and P. crinita, but within the latter he distinguished and described a distinct variety in the area of eastern Andalusia, P. crinita var. malacitana Pau, as well as three hybrid taxa: P.×P . \times composita Pau, P.×P . \times trullenquei Pau and P.×P . \times almijarensis Pau (Pau, 1918a, b, Pau, 1922) involving the three former taxa as putative parental taxa throughout the distributional range. Later, in the Iberian revision of the genus by Mateu (1986), P.lychnitis, P. crinita and four other hybrids were recognized, including P.×P . \times composita nothom. malacitana (Pau) Mateu, P.×P . \times composita nothom. composita (Pau) Mateu, P.×P . \times composita nothom. almijarensis (Pau) Mateu and P.×P . \times composita nothom. trullenquei (Pau) Mateu, without precise information about their geographical distribution. Rivas-Martínez et al. (1991) presented a quite different scenario whereby P. crinita ssp. crinita (Levante and eastern Andalusia) and P. crinita ssp. composita (Pau) RivasMartínez, Asensi, Molero & Valle (western Andalusia) hybridized to produce PP. crinita nothossp. malacitana (Pau) Rivas-Martínez, Asensi, Molero & Valle. They also suggested that Phlomis lychnitis hybridized with these two taxa to produce P.×P . \times trullenquei nothossp. trullenquei (eastern Spain) and P.×P . \times trullenquei nothossp. almijarensis Rivas-Martínez, Asensi, Molero & Valle (western and central Andalusia). On the other hand, in Málaga province (central Andalusia) Cabezudo, Nieto & Navarro (1991) considered all P. crinita-like plants to be P. crinita ssp. malacitana (Pau) Cabezudo, Nieto Caldera & Navarro. Those previously named as P.×P . \times almijarensis were reclassified as P.lychnitis, and under P. composita (pro. hybr.) they included all hybrids between P. crinita and P. lychnitis.

The present paper shows the results of a comprehensive morphometric and ecogeographical study of the P.×P . \times composita complex, formed by PP.lychnitis, P. crinita and their presumed hybrids, in the Iberian Peninsula in order to detect patterns of morphological variation within a geographical and ecological context. First, we characterized various habitats occupied by the studied taxa based on their floristic composition and environmental factors. Second, we employed multivariate ordination analyses (PCA and DFA) for assessing potential putative parental and hybrid groups, checking the existence of ecological and/or geo-
graphical patterns, if any, linked to morphological variation. Third, we applied character count procedure to determine whether intermediacy was due to divergence or hybridization.

MATERIAL AND METHODS

STUDIED PLANTS

All plants of the complex in the studied area are hemicryptophytes up to 70 cm high, representing a component of the Mediterranean vegetation in rocky limestone disturbed soils at altitudes of between 200 and 2000 m . They are tomentose with most leaves congested at the stem base, and flowers usually grouped in 6 -flowered verticillasters surrounded by two bracts. The shape of the lower leaves was the most widely employed character to distinguish between P. lychnitis and P. crinita (e.g. DeFilipps, 1972; Mateu, 1986); in the former it is linear or narrowly elliptical, gradually tapering into an indistinct petiole; in the latter it is ovate to lanceolate, cordate or truncate at base, abruptly contracted into a distinct petiole. However, all transitional leaf-shapes could also be found at individual plant level in natural populations. From a karyological point of view, the studied system is a homoploid complex since all plants are diploid (2n=20)(2 n=20) (Aparicio, 1997; Aparicio & Albaladejo, 2003).

PLANT COMMUNITIES AND ECOLOGY OF POPULATIONS

A subset of 17 populations out of 28 employed for the morphometric analysis was sampled and characterized (Fig. 1, Table 1). The composition of these communities was studied by recording all woody plant taxa in 0.1 ha plots within each population. Soil data were obtained from three pooled samples ( 0−15 cm0-15 \mathrm{~cm} cores) collected in each plot [pH, percentage of sand (S), lime (L), clay ©, CaCO3\mathrm{CaCO}_{3} (CARB), and levels of N, P, K, Ca and Mg]. Annual rainfall (AR) and mean annual temperature (T) were obtained from the nearest meteorological stations. The floristic matrix was subjected to a cluster analysis (CA) by means of Sorensen distance coefficient, using UPGMA. A canonical correspondence analysis (CCA) (Ter Braak, 1986) was then conducted on both the floristic and environmental data matrix to explore direct relationships between results of the cluster analysis and environmental variables. CA and CCA analyses were computed by PC-ORD 3.05 (McCune & Mefford, 1997).

MORPHOMETRIC ANALYSES

In total, 154 individual plants collected from 28 populations (Fig. 1, Table 1) were identified according to the model in Mateu (1986), which mainly relies on leaf

Figure 1. Location of the studied Phlomis populations. Populations analysed for ( ⊙\odot ) morphometric parameters, ( ⊙\odot ) morphometric, floristic and ecological parameters. Geographic areas supported by CA (western Andalusia, central/eastern Andalusia and Levante) are encircled.
(length, width, base shape), bract (width, colour), calyx (vein number) and trichome morphologies (see Fig. 2) to distinguish between taxa.

In our study, 13 quantitative and 13 qualitative (binary) variables were measured for every specimen (see Table 2). Values for quantitative variables were the averages of three measurements. The qualitative variables were either scored for vegetative and floral characters (INF, KM, UPL) or the presence/absence of ten different types of trichomes (Fig. 2) in seven different plant organs (lower leaf, bract, bracteole, stem base, stem apex, calyx and corolla). The trichome types were: simple (T) (uniseriate, 3-cellular), dendritic (D) (biseriate base and lateral ramifications), stellate (E) or glandular (G). Dendritic hairs either presented equivalent ramifications (D1) or a longer 3cellular stalk (D2). Stellate hairs were sessile (E1), pedunculate with equivalent ramifications (E2), sessile with a longer 3-cellular stalk (E3) or pedunculate with a pluricellular base (E4). Glandular hairs were simple (G1) or stellate with either a short (G3) or long (G4) glandular stalk.

In order to explore morphological variation, a PCA with a varimax rotation on the correlation matrix was performed. This was carried out with the whole set of variables (quantitative and qualitative), excluding LL, BL, KL and KTL to avoid redundancy, and trichome types E1, E2, D1, G1 and T, which were present in
each specimen. Continuous variables were logarithmic and vector transformed to allow the combination of quantitative and qualitative variables in the multivariate analysis, in view of the fact that the outcome employing this procedure generates scores between 0 and 1 (Brochmann, 1987; Gugerli, 1997). The variables LB, LW and LL/LW were also included in the analysis though normality was not achieved, as multivariate analyses are quite robust for departures from normality (Marcus, 1990).

In order to assess whether an a priori ecogeographical group could be discriminated efficiently and to obtain a classification matrix of the individuals within the groups, a discriminant function analysis (DFA) was performed with the same data matrix. The tested groups were: (I) P. crinita and P.×P . \times composita nothom. malacitana from Andalusia, (II) P. crinita from Levante, (III) morphologically intermediate plants ( P.×P . \times composita nothom. trullenquei, nothom. almijarensis and nothom. composita) from Andalusia, (IV) intermediate plants from Levante, and (V) P. lychnitis.

Character count procedure (Wilson, 1992) was performed to test whether the observed morphological intermediacy in the studied system was due to reticulation or divergence, comparing the assigned putative parental groups I/II and V. This method is based on selection of characters showing significant differences between the supposed parental groups, followed

Table 1. List of Phlomis populations and number of collected individuals used for morphometric analysis. Identification of individuals according to Mateu (1986). Asterisks indicate populations also subjected to floristic and ecological analyses. A, P. crinita; B, P. ×composita nothom. malacitana; C, P. ×composita nothom. composita; D, P. ×composita nothom. almijarensis; E, P. ×composita nothom. trullenquei; F, P. lychnitis

by a nonparametric one-side sign test (Zar, 1996) of intermediate vs. nonintermediate traits.

Finally, to assess the relative frequency of hybrids in mixed populations (i.e. where one or both parental and hybrid plants grow together) in the studied area, the number of individual plants morphologically regarded as extremes (parentals) and intermediates (presumed
hybrids) was recorded within every correspondent 0.1 ha plot for floristic and ecological analyses. Populations 14 and 24 were not included because only one extreme phenotype PP. crinita-like was found.

All the univariate and multivariate analyses in morphometric section, except the Dunn test, were computed using STATISTICA 5.1 (StatSoft, 1997).

Figure 2. Different types of trichomes present in the studied Phlomis taxa. Modified from Mateu (1986).

Table 2. Set of quantitative and qualitative (binary) variables assessed for morphometric analyses

RESULTS

Plant COMMUNITIES AND ECOLOGY OF POPULATIONS

CA based on floristic composition of communities showed a clear geographical pattern with three main areas: western Andalusia, central/eastern Andalusia, and eastern Spain (here called Levante sensu lato, for simplicity) (Figs 1, 3).

The results of the CCA are shown in Table 3 and Figure 4. The three geographical areas showed marked differences in a number of environmental variables (see Table 4). Populations in Levante were

Figure 3. Dendrogram of the UPGMA based on Sorensen distance coefficient for the 17 Phlomis populations included in the floristic matrix.

Table 3. Summary of results of the CCA. Only environmental variables significantly correlated with the first two canonical axes are shown ( P<0.05P<0.05 ). Abbreviations: ALT, altitude; AR, annual rainfall; T, mean annual temperature; Ca,K\mathrm{Ca}, \mathrm{K} and Mg : levels of calcium, potassium and magnesium in soil, respectively

located at the lowest altitude ( <800 m<800 \mathrm{~m} ) with the lowest values of annual rainfall, the highest mean annual temperatures, and the highest levels of Ca and K in the soil. Characteristic plant species in this area were Juniperus oxycedrus L., Thymus vulgaris L., Globularia alypum L., Satureja obovata Lag. and Genista valentina (Willd. ex Spreng.) Steud. The two western Andalusian populations analysed were at 1000 m and harboured the highest annual rainfall and Mg soil content. Representative species were Linum suffruticosum L., Thymus baeticus Boiss. ex Lacaita, Halimium atriplicifolium (Lam.) Spach and Cytisus fontanesii Spach. Finally, central/eastern Andalusian populations were at higher altitudes ( 1100−1800 m1100-1800 \mathrm{~m} ) and had the lowest mean annual temperature and K , Ca and Mg values. Characteristic species in this area were Artemisia campestris L., Rosa sicula Tratt., Teucrium hyeronymi Sennen, Prunus ramburii Boiss., Thymus membranaceus Boiss. and Astragalus granatensis Lam.

MORPHOMETRIC ANALYSES

Factor loadings and eigenvalues for the first two components (PCs1 and 2) extracted in the PCA are shown

Figure 4. CCA join-plot of studied Phlomis populations and environmental variables significantly correlated with the first two canonical axes (environmental scores ×4\times 4 ). ( △\triangle ) western Andalusia, ( Δ\mathbf{\Delta} ) central/eastern Andalusia, and ( ⊙\odot ) Levante. ALT, altitude, AR, annual rainfall, T, mean annual temperature, Ca,K\mathrm{Ca}, \mathrm{K} and Mg : levels of calcium, potassium and magnesium in soil, respectively.

Table 4. Environmental values for the three geographical areas resulting from CA and CCA. Only environmental variable values significantly correlated with the first two canonical axes are shown ( P<0.05P<0.05 ). NN, number of populations analysed in each area; ALT, altitude (m); AR , annual rainfall (mm); T, mean annual temperature ( ∘C{ }^{\circ} \mathrm{C} ); Ca,K\mathrm{Ca}, \mathrm{K} and Mg : levels of calcium, potassium and magnesium in soil (p.p.m), respectively

in Table 5. These accounted for 47.6%47.6 \% of the total variance. Leaf characters (LB, LW, LL/LW) and the calyx mouth (KM) character showed the highest (either positive or negative) correlations with PC1, and those related to bracts (BL/BW, BW, BrL) and the presence/ absence of trichomes E3 and G4 showed the highest correlations with PC2. The scatterplot for these two components showed that plants identified as P. crinita, P. lychnitis and P.×P . \times composita nothom. malacitana constituted the extremes of the morphological variation, whereas those identified as P.×P . \times composita nothom. trullenquei, nothom. almijarensis and nothom. composita were intermingled and occupied an intermediate position between the extremes, especially along PC1 (Fig. 5). Individuals of PP. crinita and P.×P . \times composita nothom. malacitana were placed in two distinct groups, corresponding to their geographical origin: group I was formed by both P. crinita and P. ×\times composita nothom. malacitana from Andalusia and group II by P. crinita from Levante. On the other hand, P. lychnitis (group V) did not show any ecogeographical pattern (data not shown). The morphologically intermediate taxa ( P.×P . \times composita nothom. trullenquei, nothom. almijarensis and nothom. composita) overlapped each other closely but were grouped separately according to their geographical provenance either from Andalusia or Levante for further analyses (groups III and IV, respectively).

Factor structure for the 17 variables and eigenvalues for the first two factors (DF1 and 2) extracted in the DFA are shown in Table 5. DF1 was positively correlated with LW, and negatively with LL/LW and LB, whereas DF2 was positively correlated with the presence of trichomes G4 and E3. DFA classification was 100%100 \% for individuals from groups I, II and V, whereas for groups III and IV it was 91.5%91.5 \% and 50%50 \%, respectively (Table 6). The scatter diagram reveals the same basic pattern previously found in PCA. Along DF1,

Table 5. Factor loadings in PCA and factor structure coefficients in DFA for the 17 morphological variables. CV, cumulative variance

groups I and II were extremes with regard to group V, whereas DF2 separated groups I and II (Fig. 6). Groups III and IV were intermediate along DF1 between groups I/II and V.

ANOVA results for the quantitative variables among groups are shown in Table 7. Putatively assigned parental groups I and V were statistically different for five out of 13 variables and the array of morphologically intermediate plants between them (group III) was intermediate for all five characters. Likewise, the putatively assigned parental groups II and V were statistically different for ten out of 13 characters and

Figure 5. PCA scatterplot for the first two extracted components. Polygons represent the ecogeographical groups I-V. Symbols follow nomenclature of Mateu (1986); (○) P. crinita; ( ∼\boldsymbol{\sim} ) P. ×composita nothom. malacitana; ( □\square ) P. ×composita nothom. composita; ( △\triangle ) P. ×composita nothom. almijarensis; ( Δ\boldsymbol{\Delta} ) P. ×composita nothom. trullenquei; ( ∙\boldsymbol{\bullet} ) P. lychnitis.

Table 6. Classification matrix of Phlomis individuals provided by DFA

group IV was intermediate for all ten characters. In both cases, one-side sign tests of intermediate vs. nonintermediate characters provided P<0.05P<0.05.

Figure 7 shows the frequency of hybrid plants in 15 mixed populations. Populations from Levante showed a clear lower percentage of hybrid plants (2.1%±2.5(2.1 \% \pm 2.5, mean ±\pm SD) than those from Andalusia ( 21.6%±11.021.6 \% \pm 11.0 ). On the other hand, western Andalusian populations showed higher percentages of hybrid plants (34.2%±1.7)(34.2 \% \pm 1.7) than those from central/eastern Andalusia (17.5%±9.3)(17.5 \% \pm 9.3).

DISCUSSION

The present study demonstrates the existence of mor-pho-ecogeographical patterns of variation in the Phlo-
mis complex in the Iberian Peninsula. The three regions distinguished by floristic and ecological analyses belong to different Iberian chorological provinces, which are based on historical, anthropogenic, geological and floristic differences (Rivas-Martinez, 1987). We found most of the Levante populations (Catalano-Valenciano-Provenzal and Murciano-Almeriense chorological provinces) growing in degraded stages of thermo-mediterranean holm-oak (Quercus rotundifolia) forests. Andalusian populations (Betic chorological province) were mostly found in mid- and highmountain mediterranean vegetation, primarily in degraded meso- or supra-mediterranean holm-oak forests.

Multivariate analyses (PCA and DFA) revealed the existence of four morphological groups, of which I, II and V constituted the extremes and III and IV the intermediate position between them. Groups I and II fit the ecogeographical pattern obtained by CA and CCA. Following Mateu’s taxonomic concept, group I includes individual plants identified as PP. crinita and P.×P . \times composita nothom. malacitana from Andalusia, while group II includes plants identified as P. crinita from Levante. Morphological characters distinguishing these two groups were the presence of trichomes E3 and G4 in bracts, bracteoles, stem apex and calyx in group I, as well as the presence of wider middle bracts. Univariate analyses also provided significant differences for seven out of 13 quantitative traits between groups I and II. In contrast, morphological

Figure 6. Discriminant function analysis (DFA) scatterplot for the first two extracted factors ( ⊙\odot ) group I; ( ⊙\odot ) group II; (□)(\square) group III; (△)(\triangle) group IV; (∙)(\bullet) group V.

Table 7. Mean ±\pm standard deviations and statistical analysis of the 13 quantitative variables for Phlomis groups. Sample size in parentheses

For each variable, parametric ANOVA ( ‡\ddagger ) or Kruskal-Wallis ANOVA ( †\dagger ) was performed among groups followed by Tukey and Dunn multiple comparison test, respectively. Variables BL, BL/BW and KL/KTL were log-transformed before parametric ANOVA.
Different superscripts indicate significant differences within rows.
aa { }^{\text {aa }} not significant; a P<0.01;aaa P<0.001{ }^{\text {a }}{ }^{\mathrm{P}}<0.01 ;{ }^{\text {aaa }}{ }^{\mathrm{P}}<0.001.
differences between plants from western and central/ eastern Andalusia within group I were not supported by PCA (data not shown). These results support to some extent the nomenclatural treatment in Cabezudo et al. (1991) where P. crinita-like plants in central Andalusia were treated as P. crinita ssp.
malacitana. Following this taxonomic approach group I should be identified as ssp. malacitana (Andalusia), and group II as ssp. crinita (Levante).

Group V included all plants identified as P. lychnitis, regardless of the geographical origin of the samples. In addition to the usually considered diagnostic charac-

Figure 7. Population structure for 15 mixed Phlomis populations. Black and white bars = frequency of morphologically extreme (parental) plants, grey bars = frequency of morphologically intermediate (hybrid) plants.
ters (narrow basal leaves with decurrent blade) (e.g. DeFilipps, 1972; Mateu, 1986; Ubera, 1987; De Bolòs & Vigo, 1995), the open calyx mouth, the unbranched inflorescence and the golden-coloured flowers could be used to distinguish PP. lychnitis from PP. crinita, usually with constricted calyx mouth, branched inflorescence and brownish or brick-orange corolla upper lip.

Groups I (P. crinita ssp. malacitana), II (P. crinita ssp. crinita) and V (P. lychnitis) should be considered the parental taxa of the hybrid complex. In contrast, groups III and IV included plants identified as P.×P . \times composita nothom. almijarensis, nothom. trullenquei and nothom. composita in Mateu (1986), but PCA did not support this nomenclaturally. All morphologically intermediate plants between PP. lychnitis and ssp. malacitana from Andalusia included in group III and between P. lychnitis and ssp. crinita from Levante included in group IV were highly intermingled. DFA wrongly classified four out of ten plants between groups III and IV, which were only significantly different in one out of 13 studied quantitative traits. These results led us to consider all intermediate phenotypes between PP. lychnitis and PP. crinita throughout their distribution range in the single nothospecies P.×P . \times composita.

Some authors have recently pointed out potential problems in hybridization studies with the use of morphological data (Rieseberg, 1995; Brochmann, Borgen & Stabbetorp, 2000). Nevertheless, morphological intermediacy is generally considered the first evidence of hybridization (Gottlieb, 1972; McDade, 1990, 1995; Dibble et al., 1998; Hawkins et al., 1999). Although the character count procedure is very conservative in its assumptions (Levin, 1999), in the present case it has given support for the hybridization hypothesis between ssp. malacitana and P. lychnitis in Andalusia,
and between ssp. crinita and P. lychnitis in Levante. In other studies the results provided by this method have been confirmed with other data sources such as in Parkinsonia L. where the hybridization hypothesis supported by the character count procedure was also confirmed by ecological, cytogenetic and pollen viability data (Hawkins et al., 1999).

Furthermore, it appears noteworthy to point out the substantial biased hybrid plant frequencies in mixed populations between Levante and Andalusia. It remains to be investigated whether this bias is due to the existence of stronger reproductive barriers between P. lychnitis and ssp. crinita than between P. lychnitis and ssp. malacitana (at either the pre- or postzygotic stage). Mixed populations in Levante apparently corresponded to bimodal hybrid zones (i.e. hybrid zones with a higher frequency of parental than hybrid individuals) where barriers against hybridization should be mainly found at a prezygotic stage (Jiggins & Mallet, 2000). Accordingly, preliminary results of a reproductive biology study involving artificial crosses showed higher seed production in handcrosses between PP. lychnitis and ssp. malacitana compared to those between PP. lychnitis and ssp. crinita (R. G. Albaladejo & A. Aparicio, unpubl. data). Furthermore, the analyses of chromosome number and meiotic behaviour in this hybrid complex showed that meiotic abnormalities were markedly more frequent in plants from Andalusia than from Levante (Aparicio & Albaladejo, 2003). This was interpreted in terms of higher levels of gene flow in Andalusian populations, which can be indicative of extensive introgression and backcrossing. Systems in which hybridization and introgression are frequent and appear to be associated with great morphological variation represent multigenerational crosses between parental taxa (Grant,

KEY FOR IDENTIFYING THREE PARENTAL TAXA

1. Lower leaves linear or narrowly elliptical up to 12 mm wide, tapering into an indistinct petiole; inflorescence usually unbranched; calyx mouth not constricted; corolla golden-yellow . . . . . . . . . . . P. lychnitis (Iberian Peninsula and southern France, 200-2000 m)

• Lower leaves ovate or lanceolate, truncate or cordate, abruptly contracted into a distinct petiole; inflorescence branched; calyx mouth usually constricted; corolla brownish to yellow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Gottlieb LD. 1972. Levels of confidence in the analysis of hybridization in plants. Annals of the Missouri Botanical Garden 59: 315-328.
Grant V. 1981. Plant speciation, 2nd edn. New York: Columbia University Press.
Gugerli F. 1997. Hybridization of Saxifraga oppositifolia and S. biflora (Saxifragaceae) in a mixed alpine population. Plant Systematics and Evolution 207: 255-272.
Hawkins JA, Olascoaga LW, Hughes CE, Contreras Jiménez JR, Mercado Ruaro P. 1999. Investigation and documentation of hybridization between Parkinsonia aculeata and Cercidium praecox (Leguminosae: Cesalpinoideae). Plant Systematics and Evolution 216: 49-68.
Jiggins CD, Mallet J. 2000. Bimodal hybrid zones and speciation. Trends in Ecology and Evolution 15: 250-255.
Levin GA. 1999. Evolution in the Acalypha gracilens/monococca complex (Euphorbiaceae): morphological analysis. Systematic Botany 23: 269-287.
Marcus LF. 1990. Traditional morphometrics. In: Rohlf FJ, Bookstein FL, eds. Proceedings of the Michigan morphometrics workshop. Ann Arbor, MI, 77-122.
Mateu I. 1986. Revisión del género Phlomis L. (Labiatae) en la Península Ibérica e Islas Baleares. Acta Botanica Malacitana 11: 177-204.
McCarthy EM, Asmussen MA, Anderson WW. 1995. A theoretical assessment of recombinational speciation. Heredity 74: 502−509502-509.
McCune B, Mefford MJ. 1997. Multivariate analyses of ecological data, Version 3.05. Gleneden Beach, Oregon: MJM.
McDade L. 1990. Hybrids and phylogenetic systematics I. Patterns of character expression in hybrids and their implications for cladistic analysis. Evolution 44: 16851700.

McDade L. 1995. Hybridization and phylogenetics. In: Hoch PC, Stephenson AG, eds. Experimental and molecular approaches to plant biosystematics. St Louis: Missouri Botanical Garden, 305-331.
Miller JT, Spooner DM. 1996. Introgression of Solanum chacoense (Solanum sect. Petota): upland populations reexamined. Systematic Botany 21: 461-475.
Pau C. 1918a. Plantas de Melilla. Boletín de la Sociedad Aragonesa de Ciencias Naturales 17: 123-133.
Pau C. 1918b. Una ligera visita botánica a Tous. Bulletin de la Institucio Catalana de Historia Natural 18: 158-161.
Pau C. 1922. Nueva contribución al estudio de la flora de Alm-
ería. Memorias del Museo de Ciencias Naturales de Barcelona Serie Botánica 1 (1): 64.
Pau C. 1925. Contribución a la flora española. Plantas de Almería. Memorias del Museo de Ciencias Naturales de Barcelona Serie Botánica. 1 (3): 6-34.
Rieseberg LH. 1995. The role of hybridization in evolution: old wine in new skins. American Journal of Botany 82: 944953.

Rieseberg LH, Carter R, Zona S. 1990. Molecular tests of the hypothesized hybrid origin of two diploid Helianthus species (Asteraceae). Evolution 44: 1489-1511.
Rieseberg LH, Ellstrand NC. 1993. What can molecular and morphological markers tell us about plant hybridization?. Critical Reviews in Plant Sciences 12: 213-241.
Rieseberg LH, Morefield JD. 1995. Character expression, phylogenetic reconstruction, and the detection of reticulate evolution. In: Hoch PC, Stephenson AG, eds. Experimental and molecular approaches to plant biosystematics. St Louis: Missouri Botanical Garden, 333-354.
Rivas-Martínez S. 1987. Memoria Del Mapa de Series de Vegetación de España. Madrid: Icona.
Rivas-Martínez S, Asensi A, Molero J, Valle F. 1991. Endemismos vasculares de Andalucía. Rivasgodaya 6: 5-76.
StatSoft Inc. 1997. STATISTICA for Windows release 5.1. Tulsa: StatSoft.
Ter Braak CJF. 1986. Canonical correspondence analysis: a new eigenvector technique for multivariate direct gradient analysis. Ecology 67: 1167-1179.
Thébaud C, Abbott RJ. 1995. Characterization of invasive Conyza species (Asteraceae) in Europe: quantitative trait and isozyme analysis. American Journal of Botany 82: 360368.

Thorpe RS. 1984. Primary and secondary transition zones in speciation and population differentiation: a phylogenetic analysis of range expansion. Evolution 38: 233-243.
Ubera J. 1987. Phlomis L. In: Valdés B, Talavera S, Fernán-dez-Galiano E, eds. Flora vascular de Andalucía Occidental Volumen 2. Barcelona: Ketres, 411-413.
Wagner WH. 1983. Reticulistics: the recognition of hybrids and their role in cladistics and classification. In: Platnick NI, Funk VA, eds. Advances in cladistics, 2. New York: Columbia University Press, 63-79.
Wilson P. 1992. On inferring hybridity from morphological intermediacy. Taxon 41: 11-23.
Zar JH. 1996. Biostatistical analysis, 3rd edn. New Jersey: Prentice Hall International Inc.