Floral primordium initiated in the axilla of bract (Figure 1(a), Figure 1(b), Figure 2(a)). Sepals of two species initiated in 2/5 helix. They initiated to heel bractlet, which also initiated in 2/5 helix; and there is no gape between the initial of sepal and bractlet. First bractlet can initiated at right or left of floral primordium (Figure 1(c), Figure 1(d)). In two species, the initiation of bractlet is slightly different. In Ph. zhejiangensis, bractlet is always initiated lean to abaxial part, but in Ph. esculenta, bractlet initiated at lean to abaxial or adaxial part (Figures 2(b)-(d) is at adaxial part and E is at abaxial part), they initiated in clockwise or anticlockwise. The first bractlet initiated always at the median of one side of floral primordium, the second one initiated at lean to adaxial part, so the first sepal initiated at non-median part of abaxial part, the exception occur at Ph. esculenta, its second bractlet occasionally initiated lean to abaxial part, so the first sepal initiated at the non-median of adaxial part. Then sepal initiated in clockwise or anticlockwise every 2/5 circle. The edge of earlier sepal is out, and the later is inner, at last formed a quincuncial perianth (Figures 1(b)-(i), Figures 2(b)-(g)). Before the sepals mature, they are evidently different each other in size, the earlier sepal is bigger than later, but when they mature, they were sub-equal in size. Two bractlets were narrow.

After perianth initiated, the centre meristem of primordium was pentagon (Figure 2(g)), at this time, the androe-

cium primordium did not initiated, after a short time, the apex meristem changed to roundness (Figure 2(h)). At this time, the edge of apex meristem changed to a ring meristem. Then some small or large protuberances formed at ring meristem, these protuberances distinct each other (Figures 3(a)-(d); Figure 2(i) and Figure 4(a)), they are androecium primordium. They did not rise in sequence; there is no sequence as sepal. Two androecium primordia could be see which have similar size, also could see single large androecium primordium. And there is no evident relationship between them and sepal. Especially the size of sepal has bigger difference in size, as the sepal growing, the first and second sepals are smaller, and the later three sepals are larger. In every flower, androecium initiated almost at the edge of primordium at same time, but they are different in growing and the difference show soon after they initiated (Figure 4(a); Figure 2(i); Figure 3(c)), at last, they trended to same size

with growing further more (Figure 4(b); Figure 3(e), Figure 3(f)). Generally stamen is 8, but it is 7 rarely (Figure 4(i)). In Ph. zhejiangensis, after the inner androecium primordium initiated, outer part continues formed androecium primordium (Figure 3(d), Figure 3(e)). The outer androecium did not initiate in sequence (Figure 3(e), Figure 3(g), Figure 3(h)), and it is 8 in most cases (Figure 5(g)). But their number may be changed sometime, so the number of androecium changed by the outer androecium in this species.

In these two species, the carpel primordium initiated following directly on its androecium in Ph. esculenta or inner androecium in Ph. zhejiangensis, and alternately with them. The primordium was separate; subsequently its basal tissue took part in carpel forming, so the carpel was united together as a whole in base (Figures 3(e)-(i);

Figures 5(a)-(d); Figure 4(b), Figure 4(c)). The primordium apex directed to outer, as gynoecium growing, it transferred to direct to top. As gynoecium primordium growing, the ring meristem of base grown at same time, and formed a ring structure which like crenelated wall. The base of gynoecium primordium developed to main part of gynoecium. The upper separate parts developed to styles and stigmas (Figure 5(e), Figure 5(f), Figure 5(h); Figures 6(a)-(d)). The basal part gradually formed ovary, which is united earlier in these two species (Figure 4(i); Figure 6(a)). In Ph. zhejiangensis the part united at all time, but in Ph. esculenta it is different, the septum between two carpel grown little, while the other part of carpel grown large, so the carpel gradually look like separate evidently. When carpel was mature, separate part was main part, so people generally thought it is separate carpel in morphology (Figures 6(b)-(e)). Carpel is 8 generally, but sometime it is 7 (Figure 5(d);

Figure 4(f), Figure 4(i)), and sometime it is 9 (Figure 5(f), Figure 5(h)).

There formed a protuberance opposite to carpel at the inner of carpel, when carpel grown to form a ring structure. This protuberance is ovule primordium (Figure 5(b); Figures 4(d)-(h)). With the carpel growing, the upper part of carpel extended to the center of floral primordium, and lay over its apex. At the same time, the septum grown to center and surround ovule. The ovule is base.

The upper part of carpel developed to style and stigma, and the stigma located at the inside of style and its face like grape (Figure 5(i); Figure 6(e), Figure 6(f)).

It is false that genus Phytolacca outer androecium opposite with sepal. This opinion was supported by Ronse

Decreane [12] . Whether to observe organogenesis or to observe its flower, there is no outer androecium opposite with sepal. Especially when the numerous of androecium is ten or much less, the androecium located in one whorl, so outer androecium which opposite with sepal did not present, and when the numerous of androecium is more than ten, the outer androecium also did not opposite with sepal. Although Ronse Decreane negated this opinion, but he still thought the androecium initiate in twist, and every two stamens opposite with every sepal. And this opinion also supported by Nishino and Chen Dan et al. [15] [16] . Accordance with our observation and the picture of other author, that androecium was thought initiated without regular relationship to sepal, but on ring meristem. Tucker (2003) first used ring meristem, she thought many taxa (29 families) have ring meristem [17] , and she definition this concept as a continued, ring, and active center non-evident meristem. The apex meristem completely accords with ring meristem, which formed androecium in Phytolacca. There is no active center; and androecium initiated irregularly. Before androecium initiated, a pause is very important for ring meris-

tem forming. This pause allows everywhere can initiate androecium on the ring meristem. The place is random where the first androecium initiated on, so that androecium located opposite or alternate with sepal just is a random phenomenon. Carpel originated follow on androecium and no any pause between them. And carpel also originated simultaneously, which originated on ring meristem evidently. The ovule originated when carpel was forming, and also originated simultaneously. So that androecium, carpel and ovule were thought originated on ring meristem, and this phenomenon maybe imply that Phytolaccaceae is one more evolutionary group. In Passiflora lobata, which is a more evolutionary group in Dilleniideae, androecium and carpel originated on ring meristem evidently [18] .

Why is the flower with 5-merous or 3-merous? This question was discussed by different author, someone think that accord with Fibonacci sequence [19] . And Nishiyama (2004) thought that 5-merous is more stable, rather than by way of Fibonacci sequence [20] . In Phytolacca, 5-merous, 8-merous and 10-merous, or more were seen, this phenomenon maybe caused by ring meristem.

In these species which have multitude stamens, the androecia developed in centrifugal sequence. Leins (1994) thought that the base group of Caryophyllales should be Cactaceae and Aizoaceae, which via Paeoniaceae connected with Magnoliidae and Dilleniidae [11] [21] . But in Phytolaccaceae, androecia were not fascicled, and have no corresponding relationship with sepal. If Cactaceae and Aizoaceae are the base group of Caryophyllales, how did Phytolaccaceae derive from them or their ancestor? The differences between Phytolacca androecium’s initiation and rotational or helix initiation in other taxa should be paid attention on. Inner rotational androecium initiated out-of-order and so do outer rotational in plants of genus Phytolacca, so it is different from other centrifugal initiation. The androecia of Phytolacca initiated on ring meristem, and because the ring meristem extended, the lower tissue of inner androecium could initiate androecium. The ring meristem initiated androecium in infinitive numerous. So that multiple numerous of androecium and centrifugal initiation all caused by ring meristem activated, and that is no same significance with other taxa which have same features.

It was proved that the carpel of Phytolacca initiated alternately with inner androecia or with androecia while single rotation [12] [22] . Is there any apocarpous pistil plant in Phytolacca? Some studies had shown there is no true apocarpous in the genus [22] [23] . There is no evident difference between apocarpous and syncarpous at initiation stage in this genus, but at its development later, because septum developed differently in different species, formed apocarpous or syncarpous, so that apocarpous and syncarpous have no same evolutionary significance in this genus and other taxa. So that couldn’t say apocarpous is original or evolutionary. These two types of carpel maybe derived parallel from their common ancestor. Scheappi (1936) thought pistil with 8 carpels is similar to Illiciaceae, so guessed that the genus has some relationship with it [24] . Endress (2001) investigated the flower organogenesis of Illiciaceae [25] , the floral initiation and development is different completely in Phytolacca and Illicia, only with same numerous of carpel. So Rohweder’s opinion [22] Phytolaccaceae has no directly relationship with Illiciaceae should be support.

After carpel initiated, ovule initiate on the inside ring meristem of carpel, this is a special character. As a structure of carpel, ovule should initiate on carpel after it to be certain size; and this is the general regulation so far. This phenomenon of ovule initiation in Phytolacca brought to mind the arguments on cauli-ovule [26] or phyllo-ovule [24] [25] [27] in it last century. Perhaps, the features of ovule initiation are helpful to treat the arguments.

Due to plants of genus Phytolacca with special characters (androecium is multiple; carpel separated and numerous), so that the systematic site of Phytolaccaceae was paid more attentions. In recent years, many studies founded different opinion on its systematic site. Manhart & Rettig (1994), Rettig et al. (1992), Downie & Palmer (1994) didn’t support the opinion which thought the family Phytolaccaceae is the basal group of Caryophyllidea [28] - [39] . Zheng et al. (2004) via to study the floral organogenesis of genus Phytolacca, thought the family Phytolaccaceae still was basal group of Caryophyllales [40] . Rodman (1994) thought family Cactaceae and Aizoaceae was the basal group, while the family Phytolaccaceae was a more evolutionary group [41] .

Many botanists expected to determine the ancestor of Caryophyllales, they speculated that it may be derived from Magnoliidea, Ranuculales or Dilleniidea, but there was no enough evidence, so did us. To investigate the androecium organogenesis of genus Phytolacca, the results show it is impossible that Caryophyllales derived from Dilleniidea or has close relationship with it. Although the androecium of Phytolacca is centrifugal, but it is different from Dilleniidea, its androecium initiated at ring meristem. Its number of stamen is much less than original group of Dilleniidea, and there is no evidently whorl or fascicle, and not evidently in sequence on time. The initiation of androecium of original plant in Dilleniidea usually divided into five districts, and the androecia united together at base in same district.