Disentangling confusions in inflorescence morphology: Patterns and diversity of reproductive shoot ramification in angiosperms

Abstract

Abstract  Terminology of inflorescence diversity has often been used in a confusing way in the literature, partly because it was based on uncritical and outdated definitions. In particular, the terms cyme, thyrse, and panicle have been misused. Although a more critical classification worked out by several authors is available, it is unfortunately not in general use because most of the relevant publications are written in German. In addition, some terms have not been used in the same way by morphologists and developmental geneticists. The present review attempts to remedy the situation with a simple outline of a classification based on: (i) different branching patterns; (ii) differential elongation of axes of different orders; and (iii) repetition of basic ramification patterns in different ways. Racemose and cymose branching are two extreme patterns; the former with limitation of axial orders to two, the second with limitation of lateral axes of each order to two. In a branching system, a sequence of racemose → cymose and, within the cyme, of dichasial → monochasial is common, but the reverse sequence generally does not occur. Systematic and evolutionary aspects of inflorescences are briefly discussed. Branching patterns are often stable in larger clades. Inflorescences of mutants studied in developmental genetic studies are mainly altered in flower or branch numbers or relative branch length, but not in branching patterns. This is also a contribution towards the goal of a unified terminology for the different fields of biology dealing with inflorescences.

Whereas flowers have been in center of interest in many branches of plant science, including morphological, developmental, functional, biological, ecological, and evolutionary aspects, inflorescences have received much less attention. Except for the studies by W. Troll and collaborators (e.g. Troll & Weber, 1953, 1955; Troll, 1957, 1964, 1969; Weberling, 1965, 1989a; Troll & Weberling, 1989; Weberling & Troll, 1998), no broad critical comparative morphological studies on inflorescences were undertaken in the past century. Because Troll's works were published in German, they did not get proper attention at a world scale. The most recent, but now rather dated, general treatments of morphological inflorescence classification by an author writing in English were provided by Rickett (1944, 1955). He used the presence or absence of terminal flowers, branching patterns, relative branch length, and the sequence of flower opening in an inflorescence as important criteria for a basic inflorescence classification. However, branching patterns and sequence of flower opening involve two very different realms of development; the first deals with initial stages of branching, the second with late differentiation and environmental influences (see below). Thus, Rickett mixed incompatible criteria in his classification. In addition, he did not provide a comprehensive and stringent classification, but remained vague in just discussing existing terms and explaining their origin and change of usage through time. This led to difficulties and confusion in the literature. Descriptions of inflorescences using such terms without good illustrations are therefore often useless. Terms that were used in an imprecise way by Rickett and other authors were “cyme”, “panicle”, and “thyrse”, and much of this confusion continues to be perpetuated even in the current literature. This is despite the much clearer definition of these terms by Troll and his collaborators, as their judgment was based on original studies on a broad range of different angiosperm groups (ca. 20 000 species studied, according to Weberling, 1983b).

Troll's (1964, 1969) classification and terminology were further elaborated and improved by Briggs & Johnson (1979), D. & U. Müller-Doblies (1987), Weberling (1989a), Weberling et al. (1993), and Rua (1999). Although the classification established and refined by these authors is almost too elaborate in some terminological details, it is currently the most practicable comparative morphological classification. A difficulty in Troll's inflorescence concept is his distinction between a “descriptive” and a “typological” (i.e. idealistic) classification. Unfortunately, the “typological” classification with two basic types, monotelic and polytelic inflorescences, appears to be of limited use, as seen from current evolutionary knowledge on inflorescences. However, his “descriptive” classification is the most comprehensive currently available. I concentrate here on the basic patterns of the descriptive classification and explain on which patterns of variation and constancy of the branching system they are based. There are various critical works that have discussed problems of subtleties and difficulties of the classification by Troll (e.g. Stauffer, 1963; van Steenis, 1963; Endress, 1970; Schroeder, 1987; Kunze, 1989; Classen-Bockhoff, 2000). Early discussions of evolutionary aspects are scarce (Zimmermann, 1935, 1965; Stebbins, 1973; Takhtajan, 1991). Troll's discussion on the exclusive direction of “transitions” from monotelic to polytelic is typological, not evolutionary. However, it should also be noted that in Troll's time evolutionary directions were difficult to establish because molecular phylogenetic analyses were not yet available.

A useful inflorescence classification for comparative morphology and macrosystematics should empirically recognize forms that are common in nature, relatively stable in (larger) clades, and logically based on a clearly established set of variables. The purpose of this review is to present the most salient traits in the diversity of inflorescences in as simple a way as possible and to discuss some evolutionary aspects.

1 Basics of growth and ramification in inflorescences

Shoots of flowering plants tend first to develop in length to conquer free space to expose the leaves and then to produce and expose flowers and fruits. These processes require elongation and branching. Branching follows regular patterns. One of these is that a new branch is formed as a rule in the axil of a leaf (phyllome), either a foliage leaf or a bract. This leaf is called the subtending leaf or pherophyll of the new branch (from the Greek “pherein” (to carry, to bear) because the pherophyll bears a lateral branch in its axil; term introduced by Briggs & Johnson, 1979). Thus, a pherophyll is defined by its position, not its shape. Pherophylls are not restricted to inflorescences, but are of general occurrence in a ramifying flowering plant. This regular branching process with a pherophyll at each branching point results in a concatenation of branches of different branching orders. The first two phyllomes (one in monocots) on each lateral branch commonly remain small and have the shape of bracts. They are called prophylls. The term “bracteoles” is also used for them in the literature, but often in a less precise manner. Such prophylls, in turn, can bear a lateral shoot in their axil. They are then at the same time prophylls of a branch and pherophylls of a branch of the next higher order in the branching system.

In inflorescences, pherophylls are more often bracts than foliage leaves. However, not every bract must have a flower in its axil, because an initiated axillary bud may not develop further. Consequently, an inflorescence may have fewer branches than there are bracts. Conversely, in some cases the pherophyll of a flower can be so reduced that it is seemingly missing, although it can sometimes be found as a rudiment in early developmental stages (e.g. flower-subtending bracts in inflorescences of Brassicaceae; Hagemann, 1963) and may reappear more conspicuously in mutants (PUCHI in Arabidopsis; Karim et al., 2009). In some Papilionoideae, floral prophylls are seemingly lacking; however, they are initiated and suppressed (Prenner, 2004). In grasses, the pherophylls of spikelets, rather than flowers, are seemingly missing (Vegetti & Weberling, 1996), but rudiments are present in early development (e.g. Ahmad et al., 2009). Mutants of the gene tassel sheath lead to loss of bract suppression (Whipple et al., 2010). In Araceae and some other Alismatales, floral pherophylls are lacking altogether (Buzgo, 2001). In the basal grade of extant angiosperms, floral pherophylls are lacking in Hydatellaceae (Rudall et al., 2007), some Nymphaeaceae (e.g. Endress & Doyle, 2009), male flowers of Hedyosmum (Chloranthaceae) (Endress, 1987), and perhaps Ceratophyllaceae (Endress, 2004).

Delimitation of an inflorescence can sometimes be difficult. As a first approximation, an inflorescence is an annual reproductive shoot or a reproductive shoot of a flush of growth (Troll, 1950,1964; Troll & Weber, 1953). However, there are cases in which it is difficult to apply this definition, especially in woody plants (e.g. Hamann, 1958; van Steenis, 1963; Troll, 1964; Endress, 1970;Hallé et al., 1978; Jäger, 1980; Urmi-König, 1981; Sell, 1995; Gleissner, 1999; Classen-Bockhoff, 2000). Then it is more suitable to proceed in a more “local” framework, focusing on the closest relatives, which may be easier to interpret. In any case, a focus on the closest relatives, in addition to a general angiosperm focus, is always important for an evolutionary understanding.

2 Inflorescence classification according to different parameters

Inflorescence diversity evolves by changes in a few basic parameters: (i) basic branching patterns; (ii) differential elongation of the axes (branches) of different orders; and (iii) repetition of branching patterns to produce “compound” (double, triple, multiple) patterns. Of the parameters listed above, the basic branching patterns primarily distinguish the different inflorescence “types” and, concomitantly, have led to most of the confusion in terminology.

2.1 Branching patterns: The two extreme forms, racemose and cymose, and their combination in the thyrse, and the developmental sequences racemose → cymose and dichasial → monochasial

Two basic contrasting branching patterns that occur in inflorescences are racemose and cymose. In the racemose pattern (Fig. 1), the main (first-order) axis (branch) has a variable (not limited) number of lateral (second-order) branches, but there are no higher-order branches. Thus, what is fixed is the number of branching orders (not more than two; thus, only first order and second order) and what is variable (not limited) is the number of second-order branches. In a racemose pattern, the main (first-order) axis can be terminated by a flower (closed, determinate inflorescence) or not (open, indeterminate inflorescence). In contrast, in the cymose pattern (Fig. 1), the first-order axis (branch) never has more than two second-order axes (branches) and never more than two extrafloral leaves (phyllomes). However, the second-order branches can branch one or more times in the same way. Thus, what is fixed is the number of lateral branches of each axis (not more than two) and what is variable (not limited) is the number of branching orders. In a cymose pattern, the first-order axis is commonly terminated by a flower, but there are also cases without a flower (see below). The cut-off between two and three next-order branches for a distinction between the two major branching patterns makes sense because the occurrence of one and two next-order branches is very common, whereas three are much more rare. This is because of the common presence of one or two prophylls in branching systems, which function as pherophylls for branches of the next higher order. A cymose branching complex is called a cyme. If in a cyme all branching axes have two lateral branches, it is a dichasium, whereas if all axes have only one lateral branch, it is a monochasium (Fig. 1). A cyme can also consist of dichasial and monochasial parts, in which case branching as a rule begins dichasial and ends monochasial (Fig. 1). The definition of cyme and cymose used here was introduced by Wydler (1851) and elaborated by Troll (1957, 1964) (see also D. & U. Müller-Doblies, 1987). Unfortunately, in American and British texts the words “cyme” and “cymose” are often not used in this sense, but have been vaguely defined (sometimes based on the centrifugal vs. centripetal opening of flowers or the presence vs. absence of a terminal flower; Rickett, 1944, 1955), which has caused much confusion (see above).

Figure 1. The two contrasting extreme branching patterns, racemose and cymose, and their combination in the thyrse. For the sake of simplicity only branches (without pherophylls and prophylls) are shown. In addition, flowers that terminate the branches are not shown in the two extreme branching patterns because the branches of the lower-most order may or may not terminate in a flower.

In both racemose and cymose patterns, the plant is flexible in the number of flowers it can produce per inflorescence. To increase flower number in the racemose pattern, the main axis becomes longer and produces more lateral flowers, whereas in a cymose pattern more branching orders are produced. Where do these two patterns occur? In an inflorescence, ramification usually starts in a racemose fashion to expand and expose flowers in the lowest-order branch and then second- or higher-order branches may become cymose (racemose → cymose). As a rule, it does not proceed the other way around. Thus, cymose ramification is secondary in development. The term “cyme” cannot be used for an entire inflorescence because the main axis of a flowering shoot regularly has more than two phyllomes; it can only be used for partial inflorescences (i.e. subunits of the branching system of an inflorescence; Troll, 1964, pp. 33, 102). Such an inflorescence with racemose primary branching and cymose secondary branching is a thyrse if open (i.e. if not terminated by a flower) or a thyrsoid if closed (i.e. if terminated by a flower). The term “thyrse” was first defined by de Candolle (1827) and more strictly by L. & A. Bravais (1837) and later by Troll (1964). For a thyrse or thyrsoid that has only one or two cymes so that its flower-bearing terminal part looks like a cyme, the term “brachium” (monobrachium or dibrachium) has been proposed (D. & U. Müller-Doblies, 1987). However, a brachium often occurs together with more richly branched thyrses and thus may represent one end of a spectrum of forms in a species and so a special term is not useful (e.g. Troll & Weberling, 1989, fig. 29, Saponaria pulvinaris). Thus, it can simply be called a thyrse with only one or two cymes. To call it a cyme, as unfortunately has sometimes been done, contradicts the definition of a cyme as having not more than two phyllomes on each axis order (see above). Another unfortunate use of terms is indeterminate for racemose and determinate for cymose (e.g. Prenner et al., 2009). An indeterminate axis lacks a terminal flower, whereas a determinate system has a terminal flower. However, racemes as most usefully defined can have a terminal flower (botryoids) and, conversely, cymes often lack a terminal flower on their first-order axis (e.g. in some Betulaceae and Fagaceae; Abbe, 1974; Fey & Endress, 1983). If a branching unit has not more than one or two (lateral) flowers, a distinction between racemose and cymose cannot be made unless it can be shown to be evolutionarily derived from a more richly branched ancestor in which the pattern is clear.

Rarely, there are specialized inflorescences that are difficult to categorize in the way explained in this section. For instance, the inflorescence form of Geranium (Geraniaceae), called “geranioid” by Schroeder (1987), can formally be seen as a multiple thyrse. However, it is somewhat special and, rather than searching for the “best fitting” general name, for a better understanding it would be more useful to compare these inflorescences primarily with those of other Geraniaceae that are less unusual.

There are several specialized terms for monochasial partial inflorescences depending on the three-dimensional pattern of ramification (e.g. scorpioid and helicoid cymes, cincinnus, bostryx, rhipidium). I will not deal with these forms here because they are subordinate to the main types. These forms have been reviewed comprehensively by Buys & Hilger (2003).

2.2 Relative length of primary and secondary axes in racemose inflorescences

In contrast with basic branching patterns, the classification of different relative lengths of axes (branches) shows how, in a system of two branching orders, different inflorescence shapes arise by differential elongation of axes of different orders (Fig. 2). This classification focuses purely on racemose inflorescences, whereas inflorescences with racemose and cymose subunits (thyrses) are not further subdivided based on differential axis length.

Figure 2. Most common forms of racemose inflorescences by variation of relative length of primary and secondary axes in the region of ramification: raceme (both long); spike (primary long, secondary short); umbel (primary short, secondary long); head (both short). In all four cases, variants with and without a terminal flower of the main axis can be distinguished.

In this context, if all axes (branches) of the first and second order are elongate, the inflorescence is a raceme (a botryum if open and a botryoid if closed). If the first-order axis is elongate and the second-order axes are short, the inflorescence is a spike (a stachyum if open and a stachyoid if closed). If the first-order axis is short in the branching region and the second-order axes are elongate, the inflorescence is an umbel (a sciadium if open and a sciadioid if closed). If all axes are short in the branching region, the inflorescence is a head (a capitulum or cephalium if open and a cephalioid if closed; see also D. & U. Müller-Doblies, 1987).

2.3 Repetition of basic (branching) patterns in racemose inflorescences and the case of the panicle

Repetition of basic branching patterns within an inflorescence is another pathway to higher-order branching (Fig. 3). If, in a racemose branching system, flowers are replaced with racemose partial inflorescences of the second order, the inflorescence is a double (or compound) raceme (diplobotryum) if open or, correspondingly, a diplobotryoid if closed; a diplostachyum if open or a diplostachyoid if closed etc. Even more complex multiple compound patterns are possible, such as a triple raceme, a quadruple raceme etc. Especially complex are inflorescences of Euphorbia, in which the well-known module, the cyathium, is a thyrsoid. In turn, numerous cyathia are commonly arranged in a thyrsoid. Thus, the entire complex is a compound thyrsoid (Müller-Doblies et al., 1975). Compound racemes exhibit an additional dichotomy, depending on whether all racemose partial inflorescences are lateral (homothetic compound raceme; i.e. all racemes of the same axial order) or whether there is also a terminal raceme (heterothetic compound raceme; i.e. racemes of different axial orders; Troll, 1964). Typhaceae exhibit especially complex multiple compound racemose inflorescences that have been studied in great detail by Müller-Doblies (1969, 1970).

Figure 3. Repetition of basic branching patterns in racemose inflorescences and the case of the panicle.

A special case is the panicle (panicula; see Wydler, 1851; Čelakovský, 1893; Pilger, 1921; Troll, 1964). In a paniculate ramification pattern, there is neither limitation in the number of branching orders nor in the number of flowers within one branching order. Because this pattern is in no way extreme, it is more difficult to describe than the racemose or cymose patterns. Each branch terminates in a flower. Branching is often richest at (or close to) the base of the main branch, where it can encompass several branching orders. Moving up the main branch, branching becomes successively less rich (encompassing fewer branching orders) and the uppermost lateral branch may consist of a single flower. A panicle could be described as a multiple compound botryoid with continuously decreasing flower numbers on the branches of the second-order and continuously decreasing branching orders towards the apex of the inflorescence. A paniculate pattern is in some way intermediate between a cymose and a racemose pattern. It is not cymose because each branch can have more than two lateral branches of the next higher order, and it is not racemose because it has branches of more than two branching orders. It differs from a compound botryoid because, moving up the first-order branch, the second-order branches do not abruptly change from second-order botryoids to single lateral flowers. Examples of groups with panicles are in Malvaceae-Sterculioideae, Hydrangeaceae, and Oleaceae (Troll, 1969).

3 Other inflorescence characteristics not connected with a general classification of inflorescence diversity

3.1 Diverse differentiation of the extrafloral phyllomes in the inflorescence

Extrafloral phyllomes (i.e. phyllomes outside of the flowers) in the inflorescence are important components of the ramification system because each ramification occurs in the axil of a subtending leaf (pherophyll), as explained above. Such phyllomes are commonly small and inconspicuous. Their main function is as protective organs for the young floral buds, which they initially cover (Endress, 1994). Often they remain small and ephemeral, but they may also become larger and colored so that they constitute conspicuous parts of the inflorescence and may contribute to inflorescence diversification. Extreme cases occur in pseudanthia (e.g. Classen-Bockhoff, 1990). However, instead of bracts or foliage leaves, sepals or petals may also contribute to pseudanthium formation (Classen-Bockhoff, 1990, 1992).

3.2 Accessory flowers and accessory partial inflorescences

In some plant taxa, a pherophyll may bear two flowers. One of them is usually smaller or more delayed than the other and is called an “accessory” flower (Sandt, 1925; Troll, 1969). The accessory flower is commonly positioned below or above the “main” flower (on the abaxial or adaxial side). More rarely there are two or more accessory flowers in a pherophyll axil. Examples with accessory flowers are Amborella (Amborellaceae; Endress & Igersheim, 2000), Dionaea (Droseraceae; Troll & Weberling, 1989), and a number of Gesneriaceae (Weber, 1978).

Instead of single accessory flowers, there may also be accessory groups of flowers, so-called accessory partial inflorescences. These are similar in complexity to the inflorescence part that they are associated with (Weber, 1975; Cavalcanti & Rua, 2008) and examples include Euphorbia (Euphorbiaceae; Müller-Doblies et al., 1975), Ipomopsis (Polemoniaceae), Coffea (Rubiaceae; Weberling & Troll, 1998), and Chirita (Gesneriaceae; Weber, 1975). In many cases, both accessory flowers and accessory partial inflorescences occur in the same inflorescence, such as in some Monimiaceae (Staedler & Endress, 2009), Thalictrum (Ranunculaceae), Swertia (Gentianaceae), or Fontanesia (Oleaceae; Troll, 1969).

3.3 Metatopies

In some inflorescences, a lateral axis appears displaced with regard to the pherophyll. Such metatopies arise either because the lateral axis is congenitally fused with the main axis for some distance (concaulescence) or the pherophyll is congenitally fused with the lateral axis for some distance (recaulescence; Troll, 1957). Metatopies are common in some larger clades, such as Polemoniaceae (Weberling & Troll, 1998), Solanaceae (Danert, 1958; Huber, 1980), and Cyperaceae (Vrijdaghs et al., 2010).

3.4 Coenosomes

Coenosomes are condensed branching systems that appear as a compact body. Developmentally, they result from rapid branching without axis elongation. Best-known examples of coenosomes are the compact inflorescences in some Moraceae and Urticaceae (e.g. Ficus; Bernbeck, 1932), some Cordia species (Boraginaceae; Troll, 1964; Hagemann, 1975; Uhlarz & Weberling, 1977), and the acervuli in chamaedoreoid palms (Uhl & Moore, 1978). In addition, cupules are commonly coenosomes (see below). All of these cases involve cymose branching. Wetland plants with numerous highly reduced flowers in compact inflorescences representing coenosomes are Hydatellaceae (Rudall et al., 2007) and Enhalus (Hydrocharitaceae; Troll, 1964). In Cordia, Enhalus, and Hydatellaceae, the flowers, on superficial inspection, seem to be initiated basipetally. However, this is an illusion caused by the compact development of the cymes in the thyrses (an explanation not considered by Rudall, 2010), in which the topographical direction of development differs from the morphological direction (see illustrations for Cordia in Hagemann, 1975).

3.5 Cupules

Cupules, compact and morphologically complex protective organs for fruits, commonly are a special kind of coenosome formed by sterile parts of the inflorescence. In Fagaceae (Fey & Endress, 1983; Nixon & Crepet, 1989) and Nothofagaceae (Rozefelds & Drinnan, 2002) they originate by rapid further (sterile) branching of the cymes after flower initiation in the lower-order branches and are composed of (often four) valves or represent a uniform cup by lateral fusion of the cymes. A uniform cup is also formed in Amphipterygium (Anacardiaceae; Bachelier & Endress, 2007). Other cupule-like structures, such as in Balanops (Balanopaceae), are not coenosomes, but simply consist of several bracts condensed below the fruit (Merino Sutter & Endress, 2003).

3.6 Cauliflory and ramiflory

Cauliflory and ramiflory designate the presence of inflorescences on the trunk or older branches of woody plants. Commonly, plants with these phenomena begin to produce inflorescences in the usual way from the axils of foliage leaves. However, in contrast with most other plants, the site of these inflorescences does not become inactive after abscission of the spent inflorescence branches, but remains active by the presence of resting buds, which repeatedly produce new inflorescence branches from season to season. Such perennial, long-lived “inflorescences” become situated on old branches as a plant ages. Resting (dormant) buds are not visible from the surface because they become more or less overgrown by the developing bark. It may be assumed that inflorescence buds are also produced de novo at new sites on trunks or old branches (or from very old resting buds?), especially in plants that bear inflorescences at the base of the trunk, where inflorescences are not present in younger individuals (e.g. Aristolochia arborea, Couroupita guianensis, Ficus sp.). The occurrence of non-axillary endogenous adventitious buds in some cauliflorous plants has been discussed by Pundir (1972) and Fink (1983). However, the details of the development of cauliflory and ramiflory are poorly understood because morphological analysis of compact ramification systems requires invasive (destructive) techniques that prevent repeated study of the same inflorescence. The best-studied case of cauliflory is Cercis (Leguminosae; Owens et al., 1995). Here, a series of several buds (main bud and accessory buds) develop in the axil of vegetative leaves and the accessory buds remain dormant and become active in years after the main bud to produce new inflorescences. In addition, buds in the axils of the lowermost bud scales of abscised inflorescences also produce inflorescences in later years. Cauliflory has evolved repeatedly in many eudicots and basal angiosperms, such as in Annonaceae (Fries, 1949) and Aristolochiaceae (González, 1999).

3.7 Opening sequence of flowers

The opening sequence of flowers is not always the same as the initiation sequence. The initiation sequence of flowers is strictly acropetal (for seemingly basipetal cases, see Coenosomes). This follows from the strictly acropetal development of a branching angiosperm shoot, in which each axillary bud is initiated together with its phyllome (e.g. Hagemann, 1973). In contrast, the maturation and opening sequence of flowers is too labile to be useful for a morphological inflorescence classification. The opening sequence may be influenced by external factors. For instance, flowers in an inflorescence may open first on the side exposed to the sun (e.g. Salix; Goebel, 1924). Often in long racemes or spikes, flowering does not begin with the basal-most flower but a few flowers higher up, whereas the basal-most flowers are somewhat retarded (Goebel, 1924; Troll, 1964); the same applies for the basal-most branches of panicles (Stauffer, 1963). Flower opening appears then bidirectional, beginning in the lower part of the inflorescence. If in an inflorescence with a terminal flower the flower opens before the lateral “lower” flowers, it is not correct to describe opening as basipetal because the terminal flower belongs to a lower axial order than the lateral flowers.

3.8 Reversion to the vegetative state

In some woody plants, inflorescences do not cease developing after flowering and fruiting, but their main axis reverts to the vegetative state, either in the same or in a later growth period, and may produce a new inflorescence later. A well-known example is Callistemon (Myrtaceae; Briggs & Johnson, 1979). This has been discussed by van Steenis (1963) and Briggs & Johnson (1979). The so-called inflorescence reversion in mutants of Impatiens (Tooke et al., 2005) is a different process, whereby the floral apex, and not the inflorescence apex, reverts to the vegetative state after formation of the outer floral organs. Thus, such a flower is greatly distorted, incomplete, and not functional.

3.9 Positional variation of floral properties according to position in inflorescences

Flowers with different properties within an individual plant are usually not randomly distributed on the plant, but follow regular patterns according to inflorescence ramifications. In Hamamelidaceae, Distylium and Distyliopsis, with botryoids or panicles, have bisexual flowers terminating the main axis and lower lateral axes, and male flowers terminating upper lateral axes in the ramification system (Endress, 1970, 1978, 1990). In andromonoecious Solanum species, branches of lower order in the cymose partial inflorescences tend to have bisexual flowers, whereas branches of higher order produce male flowers (Diggle, 1994, 1995, 2003). In the compound thyrses of Hernandia (Hernandiaceae), the cymes usually have three flowers: those of the first and second order are male and that of the third order is female (Endress & Lorence, 2004). In the botryoids of Buxus (Buxaceae), the terminal flower is female and the lateral flowers are male; in contrast, in Pachysandra, the lowermost lateral flower is female and all other flowers are male (von Balthazar & Endress, 2002). In the cyathia (thyrsoids) of Euphorbia, the terminal flower of the first order is female and all other flowers are male (e.g. Prenner & Rudall, 2007). In Kirkia (Kirkiaceae), the cymes of the thyrsoids exhibit rhythmic changes between functionally male and female flowers from one branch order to the next (Immelman, 1984; Bachelier & Endress, 2008). In some Asparagales and Liliales, the basal flowers produce more seeds than the upper ones, with the numbers gradually decreasing acropetally in the racemose inflorescences (Thomson, 1989). Diggle (2003) reviewed this pattern and others involving pollen and ovule production more broadly.

4 Systematic and evolutionary aspects

Inflorescences are shaped by developmental and ecological constraints. We should ask what forms are possible on developmental grounds (e.g. Coen & Nugent, 1994; Singer et al., 1999; Tucker & Grimes, 1999; Kellogg, 2000, 2007; Penin et al., 2002, 2005; Schmitz & Theres, 2005; Tooke et al., 2005; Vollbrecht et al., 2005; Benlloch et al., 2007; Bortiri & Hake, 2007; Conti & Bradley, 2007; Prusinkiewicz et al., 2007; Rebocho et al., 2008; Souer et al., 2008; Dumonceaux et al., 2009) and what forms are selected by pollination biological factors (e.g. Waddington, 1979; Waddington & Heinrich, 1979; Wyatt, 1982; Fishbein & Venable, 1996; Friedman & Harder, 2004; Harder et al., 2004; Jordan & Harder, 2006; Ishii et al., 2008; Makino, 2008; Classen-Bockhoff, 2009), or by other aspects of reproductive biology (Campbell, 1989; Diggle, 2003) or climate (Stebbins, 1973; Prusinkiewicz et al., 2007). Prusinkiewicz et al. (2007) discuss inflorescence morphospace and show which ramification patterns in inflorescences are common and which do not occur in nature. Curiously, they omit the thyrse with several lateral cymes, which is one of the most common inflorescence forms in nature (e.g. Troll & Weberling, 1989).

From the reviews of functional aspects with regard to reproductive biology and microsystematic aspects, it appears that inflorescence diversity is mainly related to the number of flowers or (partial) inflorescences, the longevity of such units, and the differential length of branches (Parameters 2 and 3, as discussed above), whereas branching patterns (Parameter 1) are much less affected. In contrast, inflorescence features related to branching patterns (Parameter 1) are of special macrosystematic and comparative morphological interest.

How did inflorescence evolution begin? What extremes were reached? In the current era of molecular phylogenetics and developmental biology, some dogmas on inflorescence evolution need to be abandoned. Parkin (1914) assumed that in ancestral angiosperms flowers were single and terminal on the shoots. Other authors regarded panicles (Pilger, 1922; Zimmermann, 1935, 1965; Takhtajan, 1991) or botryoids (Stebbins, 1973; called “leafy cymes” by him) as primitive. Troll did not imply evolutionary directions when he “derived” certain inflorescence forms from others. He thought of this as a kind of a didactic exercise to understand formal relationships. However, interestingly, he also viewed the panicle as a basic type from which the other forms could be “derived” (e.g. Weberling & Troll, 1998, p. 423 ff.). Panicles are monotelic inflorescences in their “typological” classification and thus both Troll and Weberling regarded polytelic as “derived” from monotelic. Weberling (1983a) regarded polytelic inflorescences as derived from monotelic also in an evolutionary sense, a process that he thought occurred many times in parallel. However, for him the question was open whether panicles or single flowers were ancestral.

With the present phylogenetic framework, it becomes possible to tackle the question of evolutionary directions with increasing precision. Among the basal angiosperms (ANITA grade), as currently conceived (e.g. Qiu et al., 1999; Soltis et al., 1999; Angiosperm Phylogeny Group, 2009), Amborella has (double) botryoids (Endress & Igersheim, 2000; Posluszny & Tomlinson, 2003). In Nymphaeales, Cabombaceae (Chassat, 1962) and Nymphaeaceae have racemes (Chassat, 1962; Endress & Doyle, 2009) and Hydatellaceae probably thyrses (Rudall et al., 2007). In Austrobaileyales, Austrobaileyaceae have single flowers or botryoids (Endress, 1980), Trimeniaceae have botryoids (Endress & Sampson, 1983), Illiciaceae have single flowers, and Schisandraceae single flowers or racemes (Weberling, 1988a). Panicles are completely absent in the basal-most angiosperms (ANITA grade) and single terminal flowers are rare and only present in Austrobaileya (in part), Schisandra (in part), and Illicium. This is also true if Chloranthaceae (Endress, 1987) and Ceratophyllum, which form the next step in the basal grade in some phylogenetic analyses, are added. In basal angiosperms the distinctions between (i) the presence or absence of a terminal flower, with a third state for solitary flowers, (ii) racemose and cymose (expressed in terms of partial units), and (iii) pedicellate and sessile flowers were treated as three independent characters in the phylogenetic data sets of Endress & Doyle (2009: characters 22, 23, 24) and Doyle & Endress (2010: characters 42, 43, 45). The current evidence from the distribution of inflorescence forms in basal angiosperms is far removed from earlier assumptions of single-flower “inflorescences” or panicles being evolutionarily ancestral (see also Doyle & Endress, 2000; Endress & Doyle, 2009). In addition, at the level of the angiosperms in general, examples are known in which the evolutionary direction is probably from polytelic to monotelic (or reversion to monotelic), in contrast with Troll and Weberling (as mentioned above), such as in Hamamelidaceae (Endress, 1970, 2003) and Lythraceae (Cavalcanti & Rua, 2008).

Studies on inflorescence diversity within smaller or larger clades show the diversity of forms and reveal which inflorescence forms are predominant in a group. A number of families were treated in Troll & Weberling (1989) and Weberling & Troll (1998). Studies on an order or single families have also been performed, such as in Myrtales (Weberling, 1988b), Amaranthaceae (Urmi-König, 1981; Acosta et al., 2009), Hamamelidaceae (Endress, 1970), Fabaceae (Weberling, 1989b), Bruniaceae (Classen-Bockhoff, 2000), Poaceae (Vegetti & Anton, 1995), Cyperaceae (Vrijdaghs et al., 2010), or Eriocaulaceae (Stützel, 1984). From these studies it can be seen that, for example, Papaveraceae, Berberidaceae, and Ranunculaceae are largely characterized by thyrsoids and panicles (Weberling & Troll, 1998), Hamamelidaceae by racemes or spikes (without terminal flowers) and panicles in some nested genera (Endress, 1970), Betulaceae by thyrses (Abbe, 1974), and Poaceae and Cyperaceae by compound inflorescences with spikes as units (see below; Vegetti & Anton, 1995; Vrijdaghs et al., 2010).

The following paragraphs concentrate on some families in which inflorescence structure has been studied in view of developmental genetics and focuses on aspects of potential evolutionary interest in these studies. In Poaceae, the general inflorescence structure is a compound spike. The so-called spikelets are spikes in the general inflorescence terminology. Grass inflorescences are so diverse because of the arrangement of spikelets into complex inflorescences. The spikelets can be arranged in spikes or racemes, or, again, in more complex patterns, such as forming a “panicle” of spikelets (Vegetti & Anton, 1995; Vegetti & Weberling, 1996; Doust & Kellogg, 2002; Perreta et al., 2009). Such panicles of spikelets are also common in Cyperoideae of Cyperaceae (Vrijdaghs et al., 2010). What evidently do not occur are “thyrses” of spikelets. This is consistent with the rule that a developmental sequence from cymose to racemose within an inflorescence generally does not occur (see above). The developmental genetic basis for the diversity of grass inflorescences has attracted special interest in view of the production of new varieties of cereals, one aspect being changes in branch numbers (e.g. Vollbrecht et al., 2005; Malcomber et al., 2006; Bortiri & Hake, 2007; Kellogg, 2007; Chuck et al., 2008).

In Brassicaceae, the leafy gene plays a role in the suppression of subtending floral bracts (see Section 2, Basics of growth and ramification; Shu et al., 2000; Sliwinki et al., 2007) and the production of double racemes instead of racemes (Schultz & Haughn, 1991; Shannon & Meeks-Wagner, 1993). Multiple compound and stunted racemes are seen in the presence of apetala1-1 and cauliflower1 (Bowman et al., 1993). In the vegetable cauliflower (Brassica oleracea var. botrytis) there is a homolog of the cauliflower gene of Arabidopsis (Kempin et al., 1995). Apetala1 may produce compound thyrses in Arabidopsis (Bowman et al., 1993). In terminal flower2 mutants of Arabidopsis, the inflorescence is terminated by a conglomerate of abnormal flowers; thus, there is not really a terminal flower (Alvarez et al., 1992; Weigel et al., 1992). The same appears to be the case in terminal flower1 mutants, as seen from the illustration in Bradley et al. (1997, figure 5), in contrast with the text. Shortening of pedicels and of internodes between flowers in Arabidopsis is caused by brevipedicellus and other genes (Douglas et al., 2002; Ragni et al., 2008). In triple mutants of terminal flower1, apetala1-1, and apetala2-1, a flower-like structure terminated by some stamens and a gynoecium appears at the apex of the inflorescence (Shannon & Meeks-Wagner, 1993). Orthologs of brevipedicellus of Arabidopsis in Brassica produce compact inflorescences by shortening of pedicels (Dumonceaux et al., 2009).

In Veronicaceae, in the model species Antirrhinum majus, inflorescences are racemes without a terminal flower (botrya). An inflorescence apex may become transformed into a floral apex by centroradialis, which is related to terminal flower in Arabidopsis (Bradley et al., 1997). This terminal flower is not normal, but peloric (Bradley et al., 1996; Davies et al., 2006). An irregular flower or several congested and reduced flowers are also not rare in many plants at the end of racemose inflorescences that normally lack a terminal flower (Buzgo & Endress, 2000; Sokoloff et al., 2006). Prophyll formation in Antirrhinum may be suppressed by the incomposita gene (Masiero et al., 2004; Davies et al., 2006).

In Solanaceae, inflorescences are thyrsoids with predominantly monochasial cymes (Danert, 1958). An unusual development in the inflorescences of Solanaceae may make it difficult to distinguish the cymes from a racemose pattern (Huber, 1980; Welty et al., 2007). Falsiflora, an ortholog of floricaula and leafy, produces inflorescence shoots instead of flowers in tomato (Molinero-Rosales et al., 1999; Hake, 2008).

Interestingly, changes in inflorescence structure by mutations reported by molecular developmental genetics in model species from these different families are mainly related to flower number and the relative length of axes or partial inflorescence number, whereas the basic branching pattern is mostly not affected. Thus, these plants exhibit changes that primarily affect Level 2, but not Level 1, of the inflorescence classification as given above.

In the future, detailed comparative inflorescence studies combined with fine-grained phylogenetic studies should show the evolutionary dynamics of inflorescences more clearly and reveal evolutionary directions. At present, it is obvious that, for example, evolutionary transitions between panicles and botryoids are easy, but we would like to know in which direction they go in each case. In Loteae (Fabaceae), inflorescences in basal clades consist of umbels of one whorl of flowers arranged in a raceme, mostly without a terminal umbel, whereas in more derived clades some of the umbels have more than one whorl of flowers. In the potential sister group (Sesbania or Robinia), there are racemes instead of umbels (Sokoloff et al., 2007). Another idiosyncratic feature is that these umbels are commonly monosymmetric and this appears to be a primitive feature in the tribe (Sokoloff et al., 2007).

5 Conclusions

Inflorescence diversity is characterized primarily by different ramification patterns. There are two basic contrasting extreme ramification patterns, namely racemose and cymose (the latter including dichasial and monochasial). Inflorescences are, in general, either racemose or combined racemose/cymose. Purely cymose inflorescences (i.e. inflorescences in which the first-order branch also does not have more than two extrafloral phyllomes) probably do not occur, which follows from the way a plant or a new shoot grows and ramifies. Thus, cymose patterns are restricted to partial inflorescences. Combined racemose/cymose inflorescences are thyrses. In a mixed branching system, the change in the branching patterns is generally from racemose to cymose, and in the cymose part from dichasial to monochasial, perhaps never the other way around. Between the two extremes, racemose and cymose, there is a third ramification pattern, exhibited by panicles, which are less common than the other patterns. Their ramification is neither racemose nor cymose because they lack the limitations of either of these extreme types.

A practically important goal is to use a unified terminology for inflorescences, tested by extensive comparative studies. The need for a unified terminology was also emphasized by Prenner et al. (2009) but, unfortunately, not consequently applied by them. It is unsatisfactory if morphologists and developmental geneticists use different terminologies and is especially confusing if the same terms are used for different phenomena, such as determinate vs. indeterminate for cymose vs. racemose or for grass spikelets with a fixed number of flowers vs. a variable number of flowers (e.g. Chuck et al., 2008). This is different from the usage in comparative morphology, in which “determinate” is used for inflorescences with a terminal flower (e.g. botryoid) and “indeterminate” for inflorescences without a terminal flower (e.g. botryum). A consequence of such lack of coordination is that users of these terms become confused and other terms also tend to be used in a sloppy and meaningless way. I hope that the present review will help in the critical application of a carefully devised comparative morphological inflorescence classification in view of the general goal of a more straightforward communication among different plant biologists.

Acknowledgments

Acknowledgements  Alex BERNHARD, Institute of Systematic Botany, University of Zurich, Switzerland, is acknowledged for graphic work. The author thanks Jim DOYLE, Department of Evolution and Ecology, University of California, Davis, CA, USA, Mary ENDRESS, Institute of Systematic Botany, University of Zurich, Switzerland, and two anonymous reviewers for valuable suggestions on the manuscript.