Observations

Both FAA‐pickled and dried samples are used (see appendix, table A1). The seed samples represent 11 genera and 106 species that cover all the genera and subgenera with dehiscent fruits as well as five genera (eight species) with indehiscent fruits. The funicle and ovule samples represent 11 genera (48 species) and eight genera (14 species), respectively (designated by superscripts a and b in the appendix, table A1). Descriptions for external morphology of seeds and placenta were made with a stereoscopic microscope and scanning electron microscopes (SEM). Longitudinal sections of seeds were observed with an optical microscope after conventional paraffin sectioning (see Fukuhara 1992). The samples for SEM observations were gradually dehydrated with ethanol, critical‐point dried, coated with gold, and observed with JEOL JSM25II and Hitachi S‐3150.

Descriptions and Terminology

I follow some descriptive terms used in previous studies on the fumarioid seeds (Brückner 1985; Fukuhara 1992) and in the taxonomic monograph of Lidén and Zetterlund (1997) for the shape of arils in Corydalis sect. Leonticoides. The “hilar region” (adopted from Brückner 1985) is the circumhilar region of the seed coat that differs from the peripheral region in some characters. I describe the spatial relationships of organs by assuming an arbitrary polarity. For instance, “the neighborhood of the hilum at the micropylar side” is often expressed simply as “above the hilum,” and accordingly, most of the figures are arranged so that the micropyle is placed above (exceptions: figs. 5.19, 8.27, 14.56, 14.59).

Seed‐coat surface sculpture will be mentioned superficially because it is already described in Fukuhara and Lidén (1995a) in combination with seed‐coat anatomy and also in Brückner (1985) and Oh and Kim (1988).

Reference Phylogeny

The reference phylogeny is shown in figure 19 and includes 20 terminal taxa (OTUs), 15 of which possess dehiscent fruits. The sections of Dactylicapnos and the subgenera of Corydalis are each represented as an OTU. Corydalis subg. Corydalis is very diverse and is subdivided into 34 (Lidén and Zetterlund 1997) or 40 (Wu et al. 1996) sections, but no satisfactory phylogenetic hypothesis within the subgenus could be constructed at present. Two informal subgroups of Cysticapnos and Ceratocapnos, respectively, are also each treated as a unit because they differ from each other in the seed characters examined here. The subtribes Fumariinae and Discocapninae are each used as an OTU because they are exclusively indehiscent fruited and there is strong evidence for the monophyly of each subtribe (table 1). The tree topology is based on the tree of Lidén et al. (1997) inferred from rps16 intron DNA sequences and morphological evidence (table 1). The clade that includes tribe Fumarieae sensu Lidén (1986) and Cysticapnos is called “Fumarieae clade” in this article. Eight of the 14 internal branches (branches marked by 1–5, 7, 9, and 10) are supported by molecular data with large confidence parameters (bootstrap values > 88%) and, except for a few branches, do not contradict morphological evidence (Lidén et al. 1997). Two branches (branches 6 and 8) have weaker support (63% and 78%, respectively) by molecular data, but there are complementary morphological synapomorphies (table 1; see also Lidén et al. 1997). The remaining four inner branches (11–14, all in Fumarieae clade) are supported purely by morphological evidence (table 1).

When inferring the pattern of character evolution, each character is considered unordered, and three out‐group taxa, Fumariaceae‐Hypecooideae, Papaveraceae s. str., and Pteridophyllaceae are added to the tree (figs. 20, 21; Drinnan et al. 1994; Kadereit et al. 1994; Hoot and Crane 1995; Hoot et al. 1997). Their seed morphological characters are based on Corner (1976) and Gunn (1980). A consistency index (CI; Kluge and Farris 1969) and a retention index (RI; Farris 1989) were calculated for each character to show the amount of homoplasy, using MacClade version 3.04 (Maddison and Maddison 1989). Because the calculations on the matrix as it is do not count the character changes within polymorphic OTUs and tend to underestimate homoplasy levels (Nixon and Davis 1991), before calculation on each character, polymorphic OTUs were broken into monomorphic subunits. These subunits were assumed to form a monophyletic group with no internal resolution.

Seed Size and Shape

The seeds are spheroidal and compressed along the lateral axis (4 in fig. 1a′). The micropyle‐chalaza‐hilum axis (1 in fig. 1a) of the ovules/seeds is curved (ana‐campylotropus ovules/seeds). As a result of the curvature, (1) the micropyle, the hilum, and (in arillate seeds) the arilbase are located close to each other; (2) the outlines of the seeds viewed from the lateral sides are oval, ovoid, kidney shaped, or, rarely, comma shaped; and (3) the endosperm is notched at the ventral side. The extent of the curvature varies between the taxa (fig. 1); in most taxa, the axis is strongly curved so that the angle of the endosperm curvature is less than 90° (fig. 1a, a′, c–e), while in Ichtyoselmis, Dicentra (fig. 1b), Dactylicapnos sect. Dactylicapnos, and Ceratocapnos, the curvature is weaker and the angle exceeds 90°. The longitudinal axis (2 in fig. 1a) is from 1–1.3 times as long as the dorsiventral axis (3 in fig. 1a), except in Cysticapnos parviflora and Cysticapnos cracca (fig. 13.50c), where the longitudinal axis is ca. 0.8 times as long as the dorsiventral one). The largest seeds (2.5–3 mm in diameter; Lemprocapnos, fig. 2a–b; some species of Corydalis sects. Radix‐cava, Leonticoides, fig. 7a–c; Capnogorium, fig. 7i; and Archaecapnos, fig. 9i–j) are about three times as long as the smallest ones (0.8–1 mm; some species of Corydalis sects. Fasciculatae, fig. 7h; Asterostigmata, fig. 9f; Chrysocapnos, fig. 9g; Ramoso‐sibiricae, fig. 9h; Cheilanthifoliae, fig. 11d–e; Thalictrifoliae, fig. 11f–g; and C. parviflora and C. cracca, fig. 13.50c). Most OTUs are heterogeneous in seed size (fig. 19, character a), and some sections of Corydalis also show large variations (sect. Corydalis, fig. 6).

Ovule size varies between 250 and 400 μm in length (fig. 15). The curvature at the ovule stages is not as variable across species as in mature seeds. Lemprocapnos (fig. 15.61), Ehrendorferia (fig. 15.63), and Dicentra (fig. 15.62) have similar anatropous ovules, but the curvature of the mature seeds is much stronger in the former two genera (fig. 1a, b; fig. 2a–d; fig. 4.9a–d).

Micropyle

The micropyle (exostome) is closed by the cells arranged in a horseshoe‐like row. The cells are as black and hard as the other part of the seed coat and often form a mound or a short beak (Dactylicapnos torulosa, fig. 5.19). These mounds or beaks are caducous in some seeds (fig. 10.39), but other seeds from the same specimen often retain them. In Pseudofumaria (fig. 14.55) and Ceratocapnos claviculata (fig. 14.56), the micropyle is overgrown by the hilar region–arilbase complex.

Hilum: General Description

In dehiscent‐fruited taxa, the hilum is a round, oblong, or laterally ovoid scar that shows a section of parenchyma in the center of which there are funicle bundles. The region of the seed coat surrounding the hilum often shows a detectable differentiation from the other regions in color, texture, and/or topography and is referred to as “hilar region” here. In indehiscent‐fruited taxa, the “hilum” is merely artificial (figs. 14.57–14.59) but is similar to that of dehescent‐fruited taxa in morphology. Although the hilum itself does not show much variation, the morphology of the hilar region and the arilbase often characterizes the taxa of Fumarioideae.

Aril: General Description

Lemprocapnos, Ichtyoselmis, Dicentra, Capnoides, Corydalis (except for sect. Bipapillatae), Dactylicapnos (except for D. aff. scandens), Pseudofumaria, and C. claviculata possess an aril at the ventral side of the seeds. The arils are whitish to pale yellow, glossy, and composed of enlarged cells. In Pseudofumaria (figs. 14.54a–b, 14.55) and C. claviculata (fig. 14.56), not only the raphal region but also the area surrounding the hilum participate in the formation of the aril (“raphe‐hilar region aril”). In the others, i.e., Lemprocapnos, Ichtyoselmis, Dicentra, Capnoides, Dactylicapnos, and Corydalis, the aril protrudes from a region below the hilum, which corresponds to the raphal surface of the ovule (figs. 15.61, 15.62, 15.66–15.69), and is called “raphal aril” here. In the raphal arils, the boundary between the aril and seed body may be “abrupt” or “continuous.” In the abrupt type (fig. 12.45), a whitish and soft arilbase protrudes from the flat seed‐coat regions and is clearly demarcated from the supporting seed‐coat region. In the continuous type (fig. 8.23), the boundary between the aril and the seed body is indistinct, and the arilbase forms a transition zone of color and hardness from seed coat to aril. The cross section of the arilbase may be round, oblong, kidney shaped, or crescent shaped and is often laterally or longitudinally elongated. The seed surface below the arilbase is notched in Capnoides (fig. 4.9e) and Corydalis subg. Corydalis (figs. 6, 7, 9).

The primordia of the arils are already present when the ovules are ready to accept pollen tubes. It may be a well‐circumscribed protuberance in Lemprocapnos (fig. 15.61) and Dicentra (fig. 15.62), being moundlike in the former and a longitudinal ridge in the latter, respectively. In Capnoides (fig. 15.66), Corydalis (figs. 15.67–15.69), and Pseudofumaria (fig. 15.70), the basal portion of the raphe is swollen as a whole to form the aril primordium, which therefore does not have a clear border to the other parts of the raphe. The apex of the primordium sometimes forms a lateral ridge (Corydalis sect. Incisae, figs. 15.67, 15.68; Corydalis sect. Sophorocapnos). There is no “vestigial” aril in the ovules that are to develop into the exarillate seeds of Ehrendorferia (fig. 15.63), Adlumia (fig. 15.64), and Dactylicapnos aff. scandens (fig. 15.65).

The shapes and sizes of the aril are diverse. In “saddle‐like arils,” the distal part forms an oblong, ovate, or lanceolate (sometimes shallowly lobed) blade, and the lateral parts of the blade are incurved to be more or less fitted to the seed body (fig. 5.20c). This type occurs in Dicentra, Dactylicapnos, and Corydalis and is associated with various types of arilbases. In many sections of Corydalis, the arilbase is more narrow, and the blade is tongue shaped, little broadened distally, and attached to the seed body (“linguiform arils,” fig. 9k). These two types, however, represent extremes of a continuous variation, and the division is arbitrary. Other types are mentioned in the following descriptions for each group.

Micropyle, Hilum, and Aril: Descriptions for Each Group

Lemprocapnos (fig. 2a, b; figs. 3.3, 3.4). 

There is no distinct hilar region, but the coloration of the hilum periphery is delayed relative to other regions during seed maturation; i.e., it is recognized as a less darkened portion before complete maturation. The arilbase is abrupt, neither hardened nor pigmented, and has a kidney‐shaped cross section. The aril primordium is a dome with an oblong base on the raphe of the ovule (fig. 15.61). The aril is distally broad and divided into several lobes, each of which may have from two to three distal lobes.

Dicentra (figs. 4.9a–d, 4.10). 

The ventral portion of the seed forms a flat zone with a narrowly elliptic outline, where the micropyle, the hilum, and the arilbase are placed. There is a groove along the midline of the zone below the hilum, and the arilbase is implanted in the groove. The arilbase is abrupt, longitudinally linear, and not pigmented. The aril is saddle‐like and small in Dicentra peregrina (fig. 4.9b, c), globose or irregularly elongated (reaching 0.5 to 2.5 times as long as the seed length) in Dicentra uniflora (fig. 4.9a), has finger‐like lobes in Dicentra cucullaria, and has several irregular lobes in Dicentra eximia (fig. 4.9d). The primordium is a ridgelike protuberance (fig. 15.62).

Dactylicapnos sect. Dactylicapnos (D. macrocapnos [fig. 5.12a, b; figs. 5.15, 5.16]; D. scandens, D. aff. scandens [figs. 5.12c, 5.13], and Dicentra paucinervia [figs. 5.17, 5.20a]). 

The ventral part of the seed is protruding. In D. paucinervia, it is further curved upward, rendering the seed comma shaped (fig. 5.20a). The micropyle and the hilum are separated by longitudinally elongated epidermal cells (figs. 5.15–5.17). In D. macrocapnos, D. scandens, and D. aff. scandens, there is a central longitudinal groove in the region, and the hilum and arilbase (when an aril is present) are placed in the groove. The hilar region is round and truncate in immature seeds (fig. 5.15) but is nearly engulfed by the swelling margins at maturity (fig. 5.16). The arilbase is abrupt and is longitudinally oblong, and the aril is globose and small. In D. paucinervia, the upper surface of the protrusion forms a truncate and pale‐colored hilar region. The apex of the protrusion forms the arilbase that continues to a peltate and lobed aril.

Corydalis subg. Corydalis (figs. 6–10). 

There is a somewhat truncate, pale hilar region. The rim bordering the hilar region is obscure in sects. Corydalis (fig. 8.23), Radix‐cava, Leonticoides (figs. 8.26, 8.27), Capnogorium (figs. 8.28, 8.31), and Chrysocapnos (figs. 10.35, 10.40), but in the other sections distinct. The hilar region of sect. Radix‐cava and C. nobilis (sect. Capnogorium, fig. 8.28) is shifted downward to the basal part of the terete aril. Corydalis scouleri (sect. Archaecapnos, fig. 10.41) is also deviating; the hilar region forms a straw‐colored dome, the lower edge of which continues to the broad arilbase. Except for sect. Duplotuber (fig. 8.29), the surface of the hilar region shows numerous pores that are intercellular spaces of the underlying parenchyma (fig. 10.41). The contour of the hilar region is kidney shaped, with the concavity corresponding to the micropyle and the convex border adjacent to the arilbase. Corydalis papilligera (fig. 6k) has a pair of swellings flanking the hilar region, being different from the other examined species of the subgenus. The arilbase is continuous, having a round (sect. Corydalis, fig. 8.23; sect. Radix‐cava; C. nobilis of sect. Capnogorium, fig. 8.28) to more or less laterally elongated cross section (the others). The aril shape is quite diverse. In sect. Corydalis (fig. 6), the arils may be straplike to terete, untwisted (C. lineariloba p.p., fig. 6e; C. caudata); clavate to falcate, untwisted (C. glaucescens, fig. 6a, b; C. turtschaninovii, fig. 6c, d; C. lineariloba p.p., fig. 6g; C. bracteata; C. orthoceras; C. pumila; C. schanginii); globose to cauliflower‐like, untwisted (C. lineariloba p.p., fig. 6f; C. papilligera, fig. 6k); terete, twisted (C. zetterlundii, fig. 6i, j; C. gotlandica; C. paczoskii; C. integra); or clavate to falcate, twisted (C. alexeenkoana, fig. 6h; C. vittae; C. angustifolia; C. malkensis). Section Radix‐cava and C. nobilis (fig. 7i) of sect. Capnogorium share long terete arils. In the remaining sections, linguiform arils and saddle‐like arils are most common. The former is found in C. arctica (fig. 7f) and C. emanueli of sect. Dactylotuber, C. flaccida of sect. Capnogorium, sect. Chrysocapnos (fig. 9g), sect. Ramoso‐sibiricae (fig. 9h, l, m), sect. Incisae (fig. 9k), sect. Fumarioides, and sect. Archaecapnos (fig. 9i, j), and the latter in C. buschii and C. ternata (fig. 7d) of sect. Duplotuber, sect. Fasciculatae (fig. 7h), C. benecincta (fig. 7g) of sect. Dactylotuber, sect. Asterostigmata (fig. 9a) except for C. sheareri, sect. Oocapnos (fig. 9b), sect. Mucroniferae, sect. Chinenses, sect. Latiflorae (fig. 9d, e), and sect. Hamatae (fig. 9c). Corydalis decumbens (fig. 7e, sect. Duplotuber) and C. sheareri (fig. 9f, sect. Asterostigmata) have deeply lobed aril blades and deviate from other members of their respective section. The arils of sect. Leonticoides are also saddle‐like but differ from the others in a thicker and broader blade. They are here classified into two apparently distinct types. One is the “spreading type,” in which the distal part extends downward (C. oppositifolia, fig. 7c; C. afghanica; C. diphylla; C. griffithii; C. uniflora), and the other is “fur‐cap type,” in which the distal part is recurved and forms a half‐open cap surrounding the hilar region (C. darwasica, fig. 7a, b; C. ledebouriana; C. maracandica; C. popovii). The former is different from the latter also in having a very broad arilbase skirting the seed body (fig. 8.27).

Seed‐Coat Surface

Seed‐coat surface is dark brown to black. The epidermal cells have straight margins and are mostly isodiametric. The surface of each cell is most commonly flat or slightly colliculate, and lustrous. Although no cellular hair is present, there is a minute protuberance in the center of the cell surface in Corydalis sect. Duplotuber (figs. 7d–e, 8.29), C. papilligera (sect. Corydalis, fig. 6k), and C. sheareri (sect. Asterostigmata, figs. 9f, 10.33, 10.34). A swollen cell surface topped with a slender spine characterizes Ehrendorferia (figs. 3.6–3.8). Dactylicapnos sect. Minicalcara and Corydalis sect. Sophorocapnos are variable in surface characters, each cell surface is flat (fig. 5.20b), convex (fig. 5.12b), or has a conical protrusion of various heights (figs. 5.12c, 5.20c, 11b, c). Within a seed, the convexity or protrusion is often more pronounced at the margin than at the lateral faces (figs. 12.44, 13.50d). In Rupicapnos (fig. 14.54e) and Fumaria (fig. 14.59), the middle of the cell surface is sunken, rendering the seed surface alveolate. Granular secondary sculpturing is present in Ehrendorferia chrysantha (fig. 3.8), Dactylicapnos aff. scandens (fig. 5.14), and a few species of Corydalis sect. Sophorocapnos (C. speciosa, C. pterophora, and C. heterocarpa).

Placenta

The scar of detached seeds (“seed scar”) is naturally a reflected image of the hilum; it exposes a section of parenchyma and a section of vascular bundle and is similar to the hilum in size and shape (e.g., compare fig. 16.73 with figs. 3.3, 3.4). The periphery of the seed scar is sometimes slightly brownish, but this character is often inconsistent even in the same fruit.

As in many other members of Papaverales, the placenta of Fumarioideae occupies the inner (adaxial) surface of replums. The placental epidermis is composed of elongated cells with smooth and flat to slightly convex surface.

The most common type of placenta is a narrow one with slightly convex surface (Ehrendorferia, fig. 16.74; Dicentra, figs. 16.76, 16.77; Adlumia, fig. 16.80; Dactylicapnos, figs. 16.78, 16.79; Capnoides; Corydalis p.p., figs. 17.83, 17.84, 17.86–17.88; Cysticapnos p.p., fig. 16.81; Pseudofumaria; Ceratocapnos claviculata). In some sections of Corydalis subg. Corydalis (Dactylotuber; Ramoso‐sibiricae, fig. 18b, b′; Incisae, figs. 17.90, 17.91, 18c; Archaecapnos p.p., figs. 17.89, 17.92), the placenta is slightly keeled. In Lemprocapnos (figs. 16.71, 16.72), the midline of the placenta is more broadly keeled to mimic a septum and further show irregular undulation. Corydalis subg. Chremnocapnos (figs. 17.85, 18e) also have uneven placental surface. In the above types, the arrangement of the funicles is uniseriate‐alternate (Adlumia, figs. 15.64, 16.80; Dactylicapnos sect. Minicalcara, fig. 16.79; Capnoides, fig. 15.66; Corydalis p.p., figs. 17.86, 17.87, 17.90; Cysticapnos p.p., fig. 16.81; Pseudofumaria, fig. 15.70; C. claviculata), biseriate‐subopposite (Lemprocapnos, figs. 16.71, 16.72; Ehrendorferia, fig. 16.74; Dactylicapnos sect. Dactylicapnos, fig. 16.78; Corydalis p.p., figs. 17.83–17.85, 17.88, 17.89), or from three to four seriate‐subverticillate (Dicentra, figs. 16.76, 16.77). Opposite or verticillate arrangement in the ovule stages tends to become subopposite or subverticilate in mature fruits (Lemprocapnos, fig. 15.60 in ovule and fig. 16.71 in seed stages, respectively).

The placenta of Ichtyoselmis (fig. 16.75) and C. vesicaria (fig. 16.82) is broader; the ovules/seeds are scattered there and do not form any recognizable rows.

Funicle

The funicles are low and inconspicuous in Ehrendorferia (figs. 15.63, 16.74), Ichtyoselmis (fig. 16.75), Adlumia (figs. 15.64, 16.80), Capnoides (fig. 15.66), Dactylicapnos sect. Minicalcara (fig. 16.79), and Corydalis p.p. (sect. Duplotuber, fig. 17.86; subg. Sophorocapnos, figs. 17.87, 17.88, 18d), Pseudofumaria (fig. 15.70), and C. claviculata. More prominent, knoblike funicles are present in Dicentra (figs. 15.62, 16.76, 16.77), Dactylicapnos sect. Dactylicapnos (figs. 15.65, 16.78), Corydalis p.p. (C. integra of sect. Corydalis; C. afghanica of sect. Leonticoides, fig. 18a; sect. Capnogorium, fig. 17.84; sect. Chrysocapnos; C. scouleri of sect. Archaecapnos), and Cysticapnos p.p. (fig. 16.81). Most of Corydalis sect. Corydalis (e.g., figs. 15.69, 17.83) and C. ledebouriana of sect. Leonticoides have longer (200–600 μm) and basally narrower (“trumpet‐like”) funicles. There is usually a fringe around the seed scar; thereby the funicle apex forms a flat to slightly concave “receptacle” for the seed (Ichtyoselmis [fig. 16.75] has a significantly broader fringe). In Ehrendorferia (fig. 16.74), Dactylicapnos sect. Dactylicapnos (fig. 16.78), Cysticapnos vesicaria (fig. 16.82), and Pseudofumaria, however, the seed scar has no distinct fringe. In Dactylicapnos sect. Minicalcara, this character cannot be assigned because the epidermal cells of the placenta are collapsed in the available herbarium samples (fig. 16.79). In Lemprocapnos (figs. 16.71–16.73) and Corydalis subg. Chremnocapnos (figs. 17.85, 18e), the seed scars are implanted in the concavity of the uneven placental surface that is fitted to the seeds. In Corydalis gigantea of sect. Archaecapnos (figs. 17.89, 17.92), the funicle apex bearing the seed scar forms a platelike structure, which has an obtusely triangular or pentangular outline that fits to the hilar region of the seeds (cf. fig. 17.92 with fig. 10.42). In three sections of Corydalis subg. Corydalis (Dactylotuber; Ramoso‐sibiricae, fig. 18b, b′; Incisae, figs. 17.90, 17.91, 18c), the funicle apex bearing the seed scar is hooked and slightly broadened. In Corydalis incisa (sect. Incisae), the oblique growth of the funicle is already observed at ovule stages (figs. 15.67, 15.68). In this and the previous groups, the seed scar faces either of the two valves of the fruit, and the seeds are flanked by the funicle apex and the valve (fig. 18b′).

The characters described above apply to the middle part of fruits and are sometimes different in the distal tapering parts, probably as a result of the spatial limitation. The arrangements of the placentas tend to be uniseriate rather than biseriate, and the placental protuberance less conspicuous.

Taxonomic Distribution of the Seed Morphological Characters: Generic Level

The diversity of 14 seed morphological characters of Fumarioideae (table 2) is coupled with the reference phylogeny in figure 19 and shows two trends of the character distribution.

First, when the character distributions are optimized on the tree, the 12 discrete characters show a high level of homoplasy and are not very informative for phylogenetic analysis at this level as expressed by low CIs and RIs. These indexes may still underestimate the amount of the homoplasy because homoplastic changes within polymorphic OTUs are not counted (Nixon and Davis 1991). Although seed surface sculptures are often the most rich source of phylogenetic signals derived from seed morphology (Cactaceae, Barthlott and Voit 1979; Ericaceae‐Pyroloideae, Takahashi 1993; Hydrangeaceae, Hufford 1995), they (characters h–j in table 2 and fig. 19) do not contain much phylogenetic information because polymorphisms within OTUs are frequent, and some states are parallel and scattered on the cladogram. Two quantitative characters, seed length (a) and funicle number per node (l), also show large variations within groups and overlaps among the groups, and I could not detect any significant phylogenetic patterns. When the characters are traced on the reference phylogeny (fig. 19), no internal branch is marked by unambiguous character changes except the Ichtyoselmis‐Dicentra clade, which is supported by weak seed curvature (b, 0) and the presence of raphal arils (e, 1). In both of the characters, parallel changes occurred in the subfamily, and the latter depends on morphological interpretation of the bright‐colored region below the hilum of Ehrendorferia (see below and fig. 20). Truncate hilar regions (c, 1) could support the Dactylicapnos‐Corydalis clade, which is weakly supported by molecular data (see “Material and Methods”) and the characters of stigma and nectary (table 1). However, it is equally parsimonious that this state is also the basal condition of Fumarieae clade and supports the monophyly of Dactylicapnos‐Corydalis and Fumarieae. Laterally elongated arilbases (f, 2) occur only in the Dactylicapnos‐Corydalis clade. It may be a synapomorphy of the clade, but it is equally parsimonious to assume it occurred parallel in each OTU. I prefer the former interpretation because sect. Minicalcara that has laterally elongated arilbases is possibly paraphyletic, and the sections of subg. Corydalis with round arilbase are likely to form a monophyletic group (see discussion on “Dactylicapnos” and “Corydalis” below).

Second, although seed morphology is not very informative for the phylogeny among the OTUs, it identifies most of the terminal taxa by a combination of characters. Ehrendorferia, Dicentra, Dactylicapnos sect. Dactylicapnos (except for D. paucinervia), Cysticapnos s. str., and Pseudofumaria are each supported by unique and derived hilar region topography (c). These autapomorphic seed characters not only complement the molecular and other evidence for the monophyly but also potentially contribute to the phylogenetic placement of fossil seeds. Three genera that show internal variation are further mentioned below.

Patterns in the Evolution of Seed Morphology: Aril

Nearly 98% of the species of dehiscent‐fruited Fumarioideae possess fleshy and whitish arils, which are often called elaiosomes. As far as hitherto studied, they attract ants so that the seeds are myrmecochorous (Berg 1969; Hanzawa et al. 1988; Nakanishi 1994; Lidén and Zetterlund 1997; Ohkawara et al. 1997). Except for Pseudofumaria and Ceratocapnos claviculata, they developed from a raphal protuberance (raphal arils). The comparison between the distribution of the arils and the phylogenetic hypotheses (fig. 20) suggests the pattern of the evolution of the arils as follows:

1. It is most parsimonious to assume three different origins of the raphal arils (fig. 20, open boxes) with two subsequent losses in Corydalis subg. Chremnocapnos and Dactylicapnos sect. Dactylicapnos, respectively, in combination with the monophyly of the Capnoides‐Dactylicapnos‐Corydalis group. On the one hand, this hypothesis is supported by the difference of the aril primordium between Lemprocapnos, Dicentra, and Capnoides‐Corydalis but needs to be confirmed by fuller sampling of taxa, especially Ichtyoselmis and Dactylicapnos. On the other hand, if the bright patch below the hilum of Ehrendorferia is homologous to the raphal aril, as is interpreted by Berg (1969), there is an equally parsimonious hypothesis that assumes a single origin. The patch is anatomically composed of translucent palisade cells (Berg 1969; Fukuhara and Lidén 1995a; called “organ‐pipe cells” by Berg) and similar to the raphal aril in the position. Under this hypothesis, the single origin of the raphal aril (fig. 20, solid box) is followed by a transfer to organ‐pipe cells (Ehrendorferia, fig. 20, doubled bar) and four losses (fig. 20, two solid ax heads in and within Dactylicapnos and Corydalis, respectively).

2. The raphe‐hilar region aril of Pseudofumaria and that of C. claviculata each has an independent origin in agreement with Lidén (1986). This interpretation is based on the morphological supports for the clades 13 and 14 of figure 19 (table 1) and is endorsed by the morphological difference of the arils between the two taxa (Brückner 1985).

The diversity in aril shape is still difficult to interpret here because of the extensive variation within the terminal units. Nevertheless, there are several examples where a small aril size is correlated to alternative seed dispersal mode. Within Corydalis subg. Corydalis, there is an obvious association between explosive fruit dehiscence and linguiform or saddle‐like arils that are small and appressed to the seed, probably to enhance flight distance at the cost of attraction to ants (Nakanishi 1994). Among the four examined species of sect. Dactylotuber, nonexplosive C. benecincta (fig. 7g) possesses much larger arils than three explosive species (fig. 7f). However, there are also numerous examples of size variation for which I can give no explanation (Dicentra, fig. 4.9a–d; Corydalis sect. Corydalis, fig. 6; Corydalis heterocarpa, fig. 11a, b). The variation in C. lineariloba (fig. 6e–g) indicates that the aril shape can change quickly during evolution, although the species is a taxonomically problematic group with much variation in flowers, fruits, and chromosome numbers (Lidén and Zetterlund 1997; T. Fukuhara, unpublished data).

Seed Stabilizer

The topography of the hilar region is variable among the terminal taxa but is reasonably constant within them (except for Dactylicapnos sect. Dactylicapnos, Corydalis subg. Corydalis, and subg. Chremnocapnos). Concave hilar regions into which the funicle apex intrudes are observed in four terminal taxa (figs. 19, 21): transverse notches in Ehrendorferia (figs. 2d, 3.6, 3.7) and longitudinal grooves in Dactylicapnos sect. Dactylicapnos p.p. (figs. 5.13, 5.16), Cysticapnos s. str. (fig. 13.51), and Pseudofumaria (fig. 14.55). Each is morphologically distinct from the others, and they have almost certainly originated independently (fig. 21). A possible function of this concavity is the stabilization of the seeds in fruits by grasping the funicle and also by increasing the area of attachment. In the taxa without a hilar concavity, the funicle apex forms a truncate or concave “receptacle” (except for unexamined Dactylicapnos sect. Minicalcara) either by a fringe around the seed scar (seven terminal taxa) or by a sunken seed scar (Lemprocapnos and Corydalis subg. Chremnocapnos). These characters also increase the area of attachment between the seeds and the placenta and could contribute to the stabilization of the seeds. I speculate that the complementary and exclusive distribution of these characters could be a consequence of the functional complementarity of the hilar concavity and the funicle receptacles (the exclusivism may reflect a natural spatial limitation). Seed stabilization may refrain immature seeds from being detached by a mechanical shock or may control the time of seed dispersal by retaining the seeds after fruit dehiscence, but detailed information on the dispersal ecology of the species is necessary to test the hypothesis. The parallel derivation of the longitudinal groove in Dactylicapnos and Cysticapnos is of interest because, in each genus, the species with the groove have exceptional fruit type for Fumarioideae (fleshy reddish pericarp in the former and balloon‐like fruits in the latter; Stern 1961; Lidén 1986, 1993; Fukuhara and Lidén 1995b). They also have broader fruits containing more numerous seeds than the species without groove. Berg (1969) suggested the concave hilar region of Ehrendorferia, which shows passively ballistic seed dispersal (sensu van der Pijl 1972), would be the area where the vibration of capsules are transformed into a pressure to trigger seeds. Although my results do not refute or support his suggestion, it may be argued that the stabilization of the seeds is itself adaptive for the passive ballist; it may hinder the release of the seeds until the wind vibrates the fruits strongly enough to effect large dispersal distance.

As mentioned earlier, in some Corydalis with explosive fruits, i.e., sections Dactylotuber p.p., Incisae, Ramoso‐sibiricae, and Corydalis gigantea (sect. Archaecapnos), the funicle apical plate is fitted to the hilar region and pushes the seed to the fruit valve (fig. 18b, b′). This character may participate in the expulsion of the seeds by giving them an outward tension or by settling the seed in a place proper for ejection. A similar function of the funicles in seed expulsion is debated for the “jaculators” of Acanthaceae (Kapil et al. 1980). Because sect. Archaecapnos is otherwise well defined and probably a monophyletic group (Lidén and Zetterlund 1997), this character must have originated at least twice in subg. Corydalis.

Both of the above‐mentioned cases emphasize the importance of funicle morphology, especially of the hilar regions, in the functional interpretation of seed morphology.