NEOTENIC EVOLUTION OF DELPHINIUM NUDICAULE (RANUNCULACEAE): A HUMMINGBIRD‐POLLINATED LARKSPUR

Takhtajan (1972) contended that many morphological discontinuities among related plant taxa are illusory because we commonly compare only adult forms. He suggested that we will discover a greater continuity of form within lineages if we consider ontogenies as well. The application of comparative ontogenetic data to questions of animal phylogeny has a long and sometimes tumultuous history (for review see Gould, 1977). Botanists, however, have only rarely dealt with broad theoretical aspects of the relationship of ontogeny and phylogeny in plants (Takhtajan, 1943, 1969 for review, 1972, 1976; Doyle, 1978). In this study, ontogenetic data together with phylogenetic and ecological information are interpreted in the context of models of ontogeny and phylogeny recently developed by zoologists, to address the evolution of the unique floral form of Delphinium nudicaule. Gould (1977) rekindled interest in the study of ontogeny and phylogeny by his thorough historical treatment of the subject. One of Gould's contributions was to distinguish between evolutionary processes (e.g., neoteny and progenesis), and their morphological or phylogenetic products (paedomorphosis). He also modified and simplified de Beer's (1930) categorization of heterochrony (i. e., changes in rates or timing of development) by reducing the basic processes to relative acceleration and retardation. Building on this foundation, Alberch et al. (1979) have developed a framework using size, shape and age as independent variables to illustrate how heterochrony can account for morphological differences among related taxa (Fig. 1). Paedomorphosis (evolutionary juvenilization), and its opposite, peramorphosis (literally 'shapes beyond'), can each be produced by three processes: paedomorphosis by neoteny, progenesis, and post-displacement; and peramorphosis by acceleration, hypermorphosis, and pre-displacement. One of the strengths of the Alberch et al. methodology is that it can be applied simultaneously to many arbitrarily chosen developmental processes within an organism. This allows different components of overall form to be conceptually dissected from one another. These can then be interpreted in the context of the whole organism. For example, Alberch and Alberch (1981) discuss the effect of several dissociations, or decouplings, of developmental processes within and among various portions of the body, on the form of the generally progenetic salamander, Bolitoglossa occidentalis. To understand the effect of heterochrony on the evolution of adult forms, it is necessary to compare the size, shape, and age to maturity of related organisms. This is problematic since maturity is not a unitary concept, but rather consists of two distinct phenomena. The ability to reproduce sexually, and a culmination of growth, are separately and together termed maturity. Because of the embryonic separation of germinal cell lines from somatic tissues in the majority of metazoans, sexual maturity in most animals is a property of whole individual organisms. An animal typically reaches sexual maturity just once, and a culmination of somatic growth is often loosely associated with the onset of sexual maturity. This is not usually the case with plants because of their indeter-

contended that many morphological discontinuities among related plant taxa are illusory because we commonly compare only adult forms. He suggested that we will discover a greater continuity of form within lineages if we consider ontogenies as well. The application of comparative ontogenetic data to questions of animal phylogeny has a long and sometimes tumultuous history (for review see Gould, 1977). Botanists, however, have only rarely dealt with broad theoretical aspects of the relationship of ontogeny and phylogeny in plants (Takhtajan, 1943(Takhtajan, , 1969(Takhtajan, for review, 1972(Takhtajan, , 1976Doyle, 1978). In this study, ontogenetic data together with phylogenetic and ecological information are interpreted in the context of models of ontogeny and phylogeny recently developed by zoologists, to address the evolution of the unique floral form of Delphinium nudicaule. Gould (1977) rekindled interest in the study of ontogeny and phylogeny by his thorough historical treatment of the subject. One of Gould's contributions was to distinguish between evolutionary processes (e.g., neoteny and progenesis), and their morphological or phylogenetic products (paedomorphosis). He also modified and simplified de Beer's (1930) categorization of heterochrony (i. e., changes in rates or timing of development) by reducing the basic processes to relative acceleration and retardation. Building on this foundation, Alberch et al. (1979) have developed a framework using size, shape and age as independent variables to illustrate how heterochrony can account for morphological differences among related taxa (Fig.  1). Paedomorphosis (evolutionary juve-nilization), and its opposite, peramorphosis (literally 'shapes beyond'), can each be produced by three processes: paedomorphosis by neoteny, progenesis, and post-displacement; and peramorphosis by acceleration, hypermorphosis, and pre-displacement.
One of the strengths of the Alberch et al. methodology is that it can be applied simultaneously to many arbitrarily chosen developmental processes within an organism. This allows different components of overall form to be conceptually dissected from one another. These can then be interpreted in the context of the whole organism. For example, Alberch and Alberch (1981) discuss the effect of several dissociations, or decouplings, of developmental processes within and among various portions of the body, on the form of the generally progenetic salamander, Bolitoglossa occidentalis.
To understand the effect of heterochrony on the evolution of adult forms, it is necessary to compare the size, shape, and age to maturity of related organisms. This is problematic since maturity is not a unitary concept, but rather consists of two distinct phenomena. The ability to reproduce sexually, and a culmination of growth, are separately and together termed maturity. Because of the embryonic separation of germinal cell lines from somatic tissues in the majority of metazoans, sexual maturity in most animals is a property of whole individual organisms. An animal typically reaches sexual maturity just once, and a culmination of somatic growth is often loosely associated with the onset of sexual maturity. This is not usually the case with plants because of their indeter-minate growth form. A single plant can be viewed as a "population" of semi-autonomous vegetative meristems, of which any can potentially become reproductive. Disregarding the problems of what constitutes an individual (ramet vs. genet;Harper, 1977), the shoot component of flowering plants has at least three levels of organization: (1) individual determinate structures, i.e., leaves and flowers (provisionally excepting the gynoecium, which may subsequently develop into a fruit); (2) individual axes, i.e., a stem and its leaves; and (3) systems of axes, i.e., the whole shoot component of a plant or parts of it. One or both notions of maturity can be applied within each level of organization. As is true of hierarchies in general, higher levels of organization of the flowering plant body have emergent properties not predicted by lower levels. It is necessary to accommodate these and other differences between plants and animals in order to use the Alberch et al. methodology to investigate morphological evolution in plants.
To facilitate ontogenetic comparison among flowers of different species, the organization of an individual flower may be compared conceptually to that of a whole animal. Both exhibit determinate growth and have distinct reproductive and nonreproductive structures. Insofar as somatic tissues of animals can be seen as analogous to the sporophytic tissues of flowers, the germinal tissues of animals have megaand microgametophytes as their floral analogs. The analogy is meant only to illustrate how both definitions of maturity can be applied to flowers; beyond this nothing is implied. That flowers are appropriate subjects for study by this methodology is a conceptual cornerstone of this investigation.
Delphinium is a large circum boreal genus noted for its strongly zygomorphic, generally blue-purple flowers specialized for pollination by bumblebees (proctor and Yeo, 1972;Faegri and van der Pijl, 1979). In California there are two red-flowered species in separate taxonomic series (Ewan, 1945), both of which are considered not 01,-Time... FIG. 1. Diagrammatic representation of the effect of heterochrony on morphological evolution. The ordinate is a non-dimensional shape axis traversed during ontogenies of paedomorphic (triangles) and peramorphic (squares) descendants relative to their ancestors (circles). The symbol a represents that age of onset of an arbitrarily chosen developmental process or event, and {j, its age of offset or culmination. The rates at which it proceeds (designated K) are indicated by the slopes of the lines. The abscissa is time, with a and {j indicated only for the ancestral condition in each case. All else being equal, paedomorphic descendants, for example, can be produced in three different ways: either (A) by advancing the age of (+8a), or (B) decreasing the age of {j (-8{j); or (C) reducing the K (-8K). These are termed postdisplacement, progenesis, and neoteny, respectively. Peramorphic descendants are produced by the opposite heterochronic changes (-8a, +8{j, or +8K), and are termed (A) pre-displacement, (B) hypermorphosis and (C) acceleration. These represent 'pure' cases where only one parameter is affected, and do not necessarily, or even usually, occur in isolation throughout an organism. Figure is adapted from several in Alberch et al. (1979). only to be specialized for pollination by hummingbirds, but also to be evolutionarily derived from bumblebee-pollinated ancestors (Grant and Grant, 1968). One of these, Delphinium nudicaule, has strongly tubular flowers that are very different in appearance from those of other larkspurs. However, they bear a striking resemblance to the buds of many other Delphinium species, for example D. decorum (Fig. 2). It is thus possible that morphological differences associated with the adaptive switch from bumblebee-to hummingbird-pollination originated as features of "juvenile" (buds) ancestors that became incorporated into the "adults" (flowers) of the derived form. If so, then D. nudicaule flowers can be considered paedomorphic.
This report has two primary goals. The first is to demonstrate quantitatively that D. nudicaule flowers are more similar to the buds than to the flowers of other, gen- eralized blue or purple-flowered larkspurs of which D. decorum is a representative. The second goal is to discriminate among three alternative processes that can produce such paedomorphic descendants. Neoteny, progenesis, and post-displacement are regarded as competing explanatory hypotheses (Fig. 1), even though it is recognized that combinations of these and other processes cannot be ruled out a priori.

MATERIALS AND METHODS
Plants from two populations of each species were sampled in spring 1979 from southern Sonoma County, California. Both populations of D. decorum, on Bodega Head and in Cheney Gulch, are in the vicinity of the type locality. The range of D. decorum extends along the California coast from Monterey to Humboldt counties, and is entirely within the range of D. nudicaule. Dephinium nudicaule was sampled in Coleman Valley and in Crane Canyon. These are both near the middle of its range, which extends from San Luis Obispo County, California in the south, to Josephine and Curry counties of southern Oregon in the north (Ewan, 1945).
The interval of time between meiosis and anthesis of the most mature anther was inferred from logistic growth curves generated both from potted plants grown under relatively uniform conditions in Berkeley, and from naturally occuring plants in the field. Sepal spur lengths were measured daily on the potted plants in Berkeley, and two to four times per week in the field. The best fit logistic curve was fitted to the data for both potted and field grown plants of each species (Appendix).
To determine the sizes of buds at which meiosis was occurring in the pollen mother cells of the largest anthers, buds were fixed in FAA, stained with acetocarmine (Radford et al., 1974), dissected and examined with a microscope (D. decorum, N = 14; D. nudicaule, N = 17). These and the spur lengths on the first day of anthesis (N = 19 for both species) were used to generate the relative times at which meiosis and anthesis occurred. To facilitate comparison between groups the average time at meiosis of each group was set at time zero. Advantages of this method are that no absolute time zero is necessary, and that the only data required are a series of lengths and the intervals of time between consecutive measurements. A drawback is that although the relative times are accurate, the absolute times for specific sizes are meaningless.
Measurements were taken of 20 floral characters on 30 open flowers and 22 characters on 30 buds, 15 from each population, of each species collected in April 1979 (Figs. 3,4). Floral measurements consisted of lengths, angles, and ordered multistate information. Except for angle of flower openness, which was measured in the field, all measurements were made on material fixed in FAA. Times were calculated for each of the buds from the logistic curves of the field grown plants, and  (Guerrant, 1978). added to the data matrix so that growth rates of different flower parts could be estimated.
Data were subjected to discriminant function analysis with four designated groups (buds and flowers of both species) using the BMDP7M program available on the University of California CDC 6400 computer. Linear regressions and descriptive statistics were done with the SPSS package on the same computer. Slopes and intercepts were compared using the appropriate t-tests.

RESULTS
The results of the discriminate function analysis of the flowers and buds of both species demonstrate that the flowers of D. nudicaule resemble the buds of both species more closely than they do the flowers of D. decorum (Fig. 5). The eight variables used to distinguish the groups are listed in the order selected, along with the coefficients of each to the three canonical variables (Table 1).
The interval of time between meiosis and anthesis in D. nudicaule is slightly longer than that in D. decorum, both in field grown plants and in potted plants grown under relatively uniform conditions in Berkeley (Fig. 6). In the field, both D. nudicaule and D. decorum took almost twice as long to mature their pollen as they did when grown in pots in Berkeley. The relationship between the species, however, is similar under both sets of growing conditions.
Allometric analyses indicate that homologous parts of the flowers of the two species have generally similar shapes at comparable sizes during ontogeny (Table  2). For example, although the right lateral sepals are significantly longer and wider in D. decorum flowers than in D. nudicaule flowers (Table 3), the slopes and intercepts of log transformed data of the sepal lengths and widths of the buds are not  Perhaps the most striking difference between the flowers of the two species is in the lower petals (Fig. 3d, h; Table 3). The petals have distinct claw and blade portions, with a "knob" at the junction. The length and width of the claw portions of   (0) sinus depth. Angles measured to the nearest 5 degrees were: (p) upper sepal hood angle, (q) upper petal angle, lower petal (r, buds only) yaw, (s) pitch, and (t) roll angles relative to the claw axis and (u) flower openness. The (v) lower petal knob was assigned to one of nine character states relative to D. decorum adults (= 9). statistically separable (Fig. 7, Table 2). This means that for the range of bud sizes sampled the right lateral sepals of both species at any given size have similar shapes (i.e., length to width ratios). The difference in adult sizes and shapes can be attributed to the slower growth rate of D. nudicaule (Fig. 8, Table 2).
A seemingly qualitative shape difference between the mature right lateral sepals of D. nudicaule and D. decorum is that the former species has cupped sepals, while the sepals of the latter are more planar. The sepals of D. decorum in the bud become increasingly more cupped until late in ontogeny when they flatten out shortly before anthesis (Fig. 9). As a con-  The ordinate is the sepal spur length in mm, the abscissa time in days from meiosis. The circles and tirangles indicate the average time of anthesis. the lower petals cannot be separated statistically, so the difference in appearance seems to be related primarily to the difference in blade sizes. The adult blade lengths and widths have completely non-overlap-ping ranges. Although statistically separable, there is considerable overlap in shape of this structure between species during development (Table 2; Fig. 10). Once again, although the two species go through ap-   Figure 5. Note that although the regression lines are statistically separable (see Table 2), there is considerable overlap between species throughout much of the range covered by D. nudicoule.
limbs are statistically indistinguishable ( Table 2). The difference in size, and therefore overall shape (ratio of limb lengths) once again can largely be explained by the different growth rates (Table 2). However, this time it is D. nudicaule that grows faster. Nonetheless, there is a difference within the nectariferous limb that cannot be accounted for by rate differences. A regression of the nectariferous portion of the nectariferous limb ( Fig.  4j) to its total length (Fig. 4h) in buds shows both the slopes and intercepts to be different (Table 2, Fig. 12). This means that proportionately more of the nectariferous limb of D. nudicaule is given to the saccate portion, which is here assumed to be related to nectar production. At any given size, a petal of D. nudicaule will presumably be able to produce more nectar than one of D. decorum.

DISCUSSION
Before the primary goals can be discussed, two premises that underlie this study will be addressed. The first is that the floral form of D. nudicaule is likely derived from a generalized bumblebee- proximately the same series of shapes at similar sizes, the great discrepancy in adult sizes can largely be accounted for by the slower growth rate of D. nudicaule (Table 2; Fig. 11).
Not all of the features of D. nudicaule flowers resemble developmental stages of D. decorum. The upper or nectariferous petal of D. nudicaule (Fig. 3g) is not only larger than that of D. decorum (Fig. 3c), but also exhibits other differences. The allometric relationships of the nectariferous (~ig. 4h) to non-nectariferous ( Fig. 4i) 2 4 6 Sepal width FIG. 9. Plot of right lateral sepal width vs. depth for buds and flowers of both species, indicating that the sepals of both species are cupped until shortly before anthesis when only those of D. decorum become more planar. Symbols as in Figure 5.  19.548 in "states" 1-9 dec 9.0 0.0 9-9 29 *** *** = P < .OOI.
pollinated type. The second is that D. de-genera with north temperate floristic afcorum flowers are a reasonable represen-finities that are presumably originally beetative of the probable ancestral floral type. pollinated, but which also have one or a Grant and Grant (1968) note that Del-few hummingbird-pollinated species in phinium is one of 19 broadly distributed western North America. Combining this with the predominantly neotropical distribution of hummingbirds, they suggest that within these groups hummingbird pollination is derived. Both D. nudicaule and D. decorum are members of the tuberiform series, which is almost exclusively limited to extreme western North America (Ewan, 1945). In addition to D. decorum, the tuberiform series also contains 12 other bumblebee-pollinated species. All of these have flowers that are very similar not only to those of one another, but also to those of many species in practically all other series in both the New and Old World. By applying the criterion of outgroup analysis (see Stevens, 1980) at the series level (using any series with flowers that resemble either of the two species in this study as the out-group), it is reasonable to conclude that within the tuberiform series, the floral form of D. nudicaule, which is unique within the genus, is derived and that the bumblebee-pollinated type is ancestral and generalized. This conclusion is similar to that reached by Grant and Grant (1968), who state that D. nudicaule "must have diverged from a bee-flowered ancestor." Ideally, a derived taxon should be compared directly to its sister group that shows the ancestral condition. However, because the sister group of D. nudicaule is not  (Lewis et al., 1951). The genetic distance between them is likely not great; naturally occurring hybrid swarms between these two species have been reported (Munz, 1959;Howell, 1970;Santana, 1975;Guerrant, 1978), and viable seed can be obtained from experimental crosses (Santana, 1975;Guerrant, unpubl.).
A typical Delphinium flower (Figs. 2c, 4a-d) has five widely spreading, generally blue-purple, petaoloid sepals that surround the two pairs of highly modified petals which in turn conceal the stamens and pistils. The orange-red flowers of D. nudicaule (Figs. 2a, appear tubular because the sepals do not spread widely apart at anthesis, and are curved in transection. In a typical larkspur, the greatly expanded blades of the two clawed lower petals meet along the midline of the zygomorphic flower. These are used as a landing platform by bumblebees (Laverty, 1980), and apparently serve also to deny all but the heaviest and strongest flower visitors access to the nectar. In D. nudicaule, the lower petals have very small (seemingly vestigial) blades, positioned off to the side of the flower, which do not (Alberch et al., 1979; Fig. 1). Those feaphysically obstruct access to the nectar. tures that most noticeably distinguish D.
The presumed ancestral function of exnudicaule flowers from D. decorum and eluding illegitimate flower visitors by the other typical larkspur flowers are its forlower petal blades is apparently supplantward pointing cupped sepals and its small ed by the tubular form, red color, and lower petal blades. Allometric analyses indownward pointing angle at which the dicate that the lateral sepals of both species flower is positioned. These features com-grow through size-shape trajectories statisbine to seemingly make D. nudicaule nee-tically indistinguishable from one another tar relatively inaccessible to bumblebees. throughout the range of sizes measured. The initial goal of this study was to The size-shape paths of lower petal blades, quantify the relative similarity of D. nu-though statistically separable, overlap dicaule flowers to the buds and flowers of broadly during development. In other D. decorum. A discriminant function words, when these structures are the same analysis of a broad array of floral char-size they have approximately the same acters supports the hypothesis that D. nu-shapes. Furthermore, the cupped aspect of dicaule flowers are more similar to D. de-D. nudicaule sepals, and their forward corum buds than to its flowers. It follows pointing orientation in the flower, reprethat if D. nudicaule is actually derived sent conditions which in D. decorum are from a taxon with flowers similar to those seen only in the buds. The great differof D. decorum then its flowers can be con-ences in adult sizes of these features, and sidered to have become paedomorphic, or to varying degrees in shape also, can be evolutionarily juvenilized in appearance.
attributed to the conspicuously slower Assuming that D. nudicaule flowers are growth rate of D. nudicaule. Hence, negenerally paedomorphic, the next goal is oteny accounts, to a considerable degree, to discriminate among the various possible for the externally visible differences in processes by which this could have come shape that distinguish the flowers of these about. To do this, it is necessary to com-two species. pare the ages at maturity of the two taxa.
These features, along with the red col-Delphinium nudicaule required a slightly or, seem to serve both to attract humlonger period to mature its pollen both mingbirds, and make the flowers less suitunder field and uniform garden condi-able nectar sources for bumblebees. The tions. It must be emphasized I do not sug-evolution of red flower color from a bluegest that morphological differentiation of purple ancestor requires no explanation the flowers begins with meiosis. Rather, beyond a traditional view of selection by meiosis is simply a convenient, easily dis-hummingbirds acting on existing variatinguished, discrete early developmental tion. Delphinium nudicaule flowers conevent that is used as a marker. Because tain both red and blue pigment complexes both progenesis and post-displacement en-that are believed to be based respectively tail a reduction in the time required to on the anthocyanidins pelargonidin and mature structures in question, (a to f3 in-delphinidin (Guerrant, 1978). Hypothetiterval, Fig. 1), they are effectively rejected cally its immediate ancestor contained as hypotheses. Neoteny then becomes the these also, though with relatively more primary hypothesis for the process that blue than red pigment, compared to D. produced the novel floral form; and this nudicaule. was presumably selected for by humming-Though similar, D. nudicaule flowers birds.
are not identical to the buds of D. deco-If D. nudicaule flowers are of "pure" rum. This should come as no surprise neotenic origin, all sporophytic floral parts since mosaic patterns of evolution, in would then be expected to grow through which different portions of organisms the same series of shapes, though at a evolve separately, are widely acknowlslower rate, than those of D. decorum. edged (Mayr, 1963;Dobzhansky et al., 1977). Whereas the externally visible portions of the flowers seem generally neotenic, the nectariferous petals, which produce the pollinators' reward, are not. The nectariferous and non-nectariferous limbs have a comparable allometric relationship to one another in both species. However, those of D. nudicaule are larger and have a different shape (ratio of nectariferous to non-nectariferous limb lengths) since they grow faster than those of D. decorum. This can be interpreted as a peramorphic change that occurred by a combination of the processes of acceleration and hypermorphosis (Fig. 1). Within the nectariferous limb, both the slope and intercept are different between species in a regression of the nectariferous portion to the total length. This means that the nectariferous petals of D. nudicaule are not only larger in absolute terms, but proportionally more of them are devoted to nectar production. Additionally, while both the sepal spurs and nectariferous limbs of the upper petals of D. nudicaule are longer than those of D. decorum, the petals seem to have become disproportionately larger. Although the differences are not statistically significant, the average petal spur lengths are greater than the sepal sacs that contain them in D. nudicaule, while they are smaller in D. decorum. This suggests that either there were differential selective pressures on these structures, or they had different sources and magnitudes of variation associated with their origin.
Together these differences should account for the greater nectar production in D. nudicaule which is consistent with its pollination syndrome. Hummingbird-pollinated flowers characteristically produce more nectar per flower than do bumblebee-pollinated flowers (Baker and Baker, pers. comm.). The concentration and composition of sugar in the nectar of D. nudicaule are also consistent with predictions based on a large sample of nectar from flowers with many different pollination syndromes (Baker and Baker, 1979, in press). Delphinium nudicaule produces copious dilute (34% sucrose equivalents, wt. to total wt.), sucrose-dominated nectar (Guerrant, 1978). Although 34% is near the upper limit of sugar concentrations found in hummingbird-pollinated species, D. nudicaule, and D. cardinale, the other red-flowered hummingbird pollinated larkspur, have the lowest nectar sugar concentrations found in all larkspurs surveyed (Baker and Baker, 1979, in press, and pers. comm.;Guerrant, unpubl.). Many of the bumblebee-pollinated species, for example D. decorum (47%), have nectar that averages around 50% sugar.
In summary, the juvenile appearance of D. nudicaule flowers relative to those of D. decorum, can in part be accounted for, and was most likely produced by, the process of neoteny. In other words, although the flowers of both species have comparable times to maturity, those of D. nudicaule do not progress as quickly through their mutual series of shapes, so their flowers end up looking like buds of D. decorum. The resulting tubular flower shape represents convergent evolution onto a floral form that is commonly visited by hummingbirds. The ability to produce the increased nectar reward, which seems clearly associated with the switch from bumblebee to hummingbird pollination, can be attributed to a localized combination of the processes of acceleration and hypermorphosis of the nectariferous petal. In addition, the morphological trend toward greater nectar production potential has been augmented by a novel allometric reorganization of the nectariferous limb of the nectariferous petal. Implicit in this study is an opportunity to provide a first approximation evaluation of the Alberch et al. (1979) methodology at elucidating morphological evolution in plants. Perhaps the most immediately apparent value of this methodology is that it provides a clear definition of, and therefore distinction among, components of form. Using these sorts of developmental parameters in conjunction with allometry, 'null' hypotheses of expected forms can be generated. Comparing these to empirical results in an ecolog-ical and phylogenetic context may then provide further insight into the nature and role of adaptation and natural selection in evolution (see Alberch et al., 1979;Alberch, 1981).
At least in this case, the morphological changes in structures mediated by different processes correspond closely to their presumed changes in function. It can be hypothesized that if an initial neotenic event occurred, it would simultaneously have created the tubular shape characteristic of hummingbird-pollinated flowers, and rendered the nectar less accessible to bumblebees. This, in a sense, may have constituted the 'key' adaptation to hummingbird pollination. It would have provided the basic framework upon which subsequent selection by hummingbirds has fine tuned the change in pollination syndromes. Comparative developmental studies, it seems, provide a rich conceptual bridge between ecology and evolution, in a manner that approaches Van Valen's (1974) aphorism, "evolution is the control of development by ecology." Finally, although Alberch et al. did not present their methodology as a 'theory of ontogeny and phylogeny', it does provide a coherent framework, which when taken in conjunction with ecological and phylogenetic information, can be used to generate falsifiable hypotheses about the origin of novel forms and the evolution of shape in general. For example, it has been hypothesized that a reduction in the time required to flower and produce seed was a primary selective force that led to the evolution of autogamy in species of Clarkia, (Moore and Lewis, 1965), Leavenworthia (Lloyd, 1965;Solbrig and Rollins, 1977), and Limnanthes (Arroyo, 1973). If so, it is possible that the small flowers that characterize these and some other derived autogamous species may simply be a result of a progenetic origin (Fig. 1). An increased appreciation of the constraints development imposes on morphological evolution may ultimately yield insights into the origins of morphological variation, and in turn, the nature and relationship of ad-aptations and their consequences (see Gould and Lewontin, 1979). SUMMARY The flowers of the derived hummingbird-pollinated D. nudicaule resemble the buds of the generalized bumblebee-pollinated species much more closely than they do their flowers. Delphinium nudicaule flowers then may be considered paedomorphic, in that the adults (flowers) of the derived species resemble the juveniles (buds) of the ancestral type. This study brings to bear current ideas, developed by zoologists, about the relationship of ontogeny and phylogeny in an attempt to explain the processes by which the morphological differences that accompanied the evolution of D. nudicaule flowers may have originated. Comparative allometric and growth rate data suggest that relatively few developmental perturbations are required to account for the bulk of the morphological differences observed. The externally visible portions of the flower, which apparently function both to attract hummingbirds and make the nectar less accessible to bumblebees, seem generally neotenic in origin. However, the nectariferous petal, which produces the pollinator's reward, is not juvenilized. Rather, it has recapitulated the ancestral shape and become peramorphic by a combination of the processes of acceleration and hypermorphosis. It is suggested that the dissection of component processes that accompany morphological change during evolution, when used in an ecological and phylogenetic context, allows a more critical understanding of the origin of morphological variation and the nature of adaptations and their consequences. Linda E. Newstrom, Kevin Padian, Robert H. Robichaux, Peter F. Stevens, Elizabeth Taylor, and an anonymous reviewer for their critical advice on various earlier versions of this manuscript. I am grateful to Pere Alberch and Peter F. Stevens for valuable comments on conceptual aspects of this paper. Rick Berg, an economist, kindly devised the method I used to estimate, and make statistical inferences from, the logistic curve. Thomas Duncan assisted with the computer work, and arranged for funds for computer time. Finally, I especially thank David B. Wake, who, while under the impression he was talking about salamanders, explained to me a good deal of what I discuss in relation to larkspurs, and who has very patiently assisted me in all phases of the research and writing of this paper. I am grateful to all of these people for these and many other kindnesses. The California Native Plant Society provided funds, which greatly assisted my field work on the project.