Int. J. Plant Sci. 162(4):697–717. 2001.
! 2001 by The University of Chicago. All rights reserved.
1058-5893/2001/16204-0002$03.00
THE DEVELOPMENTAL BASIS OF FLORAL VARIATION IN DRIMYS WINTERI (WINTERACEAE)
Andrew N. Doust1
University of Melbourne, Parkville, Victoria 3052, Australia
Flowers within a tree of Drimys winteri vary markedly in floral organ number and arrangement. This
variation can be linked to developmental differences between flowers in different parts of the inflorescence.
Floral organ number varies most markedly between terminal and lateral flowers, whereas variation in organ
arrangement is correlated with the shape of the flower. Key events during development, including changes in
meristem shape and a delay in initiation of petals, allow a remarkable variety of whorled and spiral phyllotactic
patterns to arise. Relatively few flowers showed Fibonacci phyllotaxis, which is otherwise very common in
plant systems. Variation in divergence angle within flowers is correlated with meristem shape, whereas variation
in plastochron ratio is shown to depend both on meristem shape and on divergence angle. Differences in floral
morphology are interpreted as the result of an ordered process of initiation of primordia superimposed on
variation in the symmetry of the floral meristem. In flowers with more or less circular floral meristems, organs
appeared to be initiated sequentially at a regular divergence angle, whereas during early development of flowers
with more elliptic meristems there was a preferential initiation of organs toward the poles of the elliptic
meristems. This study is the first to examine how variation in floral organ number and arrangement is partitioned within individuals of D. winteri, or indeed any other member of the Winteraceae.
Keywords: Drimys winteri, Winteraceae, floral morphology, floral development, phyllotaxis, divergence angle,
plastochron ratio, morphological variation.
Introduction
iation in carpel arrangement originated in the first tier of carpels to be initiated. Each successive tier of carpels tended to
repeat the pattern of divergence angles found in the first tier.
Divergence angles in the carpel zone were variable and occasionally over 180", allowing the spiral comprised of sequentially initiated primordia occasionally to reverse its direction. However, the mechanism by which the pattern of
carpel arrangement in the first tier was propagated throughout
subsequently initiated primordia was not investigated.
A more comparative approach to the study of floral organization was taken by Endress (1986), who contrasted three
groups of magnoliids, (1) Degeneriaceae, Himantandraceae,
Eupomatiaceae, and Austrobaileyaceae, (2) Chloranthaceae,
Trimeniaceae, and Amborellaceae, and (3) Winteraceae. The
first group was chosen for their complicated and conspicuous
flowers with regular spiral phyllotaxy, the second for their
relatively small and inconspicuous flowers, and the third for
their unusually large variation in floral organ number and size.
Endress (1986) considered that phyllotaxy in the flowers of
Winteraceae was more or less unordered, and he linked irregularity in mature organ arrangement with the often asymmetrical floral apex and the small size of the primordia relative
to the floral apex. He concluded that the flowers of primitive
angiosperm groups could have a diversity of floral arrangements and a variable number of floral organs.
Most studies of Winteraceae have focused on Drimys winteri
J.R. and G. Forst., and substantial variation in floral organ
number and arrangement was noted by Smith (1943) and
Tucker (1959) (table 1). Vink (1970) and Erbar and Leins
(1983) observed variation in organ arrangement in the flowers
of this species and postulated that the sometimes asymmetrical
Extant flowering plant families exhibit a wide diversity of
floral form. This morphological diversity can be dissected into
variations in the number, arrangement, and morphology of the
floral organs (perianth, androecium, and gynoecium). The
number and arrangement of floral organs is the direct outcome
of the pattern of production of primordia by the floral meristem, whereas the morphology of the floral organs is the outcome of the differentiation of the primordia. A distinction between production and differentiation of primordia is important
in understanding how diversity of floral form arises. In many
families organ number and position are relatively stable and
most variation is seen in the elaboration of those organs. This
reaches an extreme in, for example, the monocot family Orchidaceae and the dicot family Fabaceae, where an almost
invariant floral architecture underlies dramatic differences in
both floral organ and overall flower morphology. However,
flowers in families among the basal lineages of the angiosperms, such as Winteraceae, often vary markedly in organ
number and arrangement but relatively little in floral organ
morphology (Endress 1986).
Variation in organ number and arrangement implies variation in the production and placement of primordia by the
floral meristem. Tucker (1960, 1961) studied ontogeny of flowers in Michelia fuscata (Magnoliaceae) and concluded that varCurrent address: University of Missouri—St. Louis, 8001 Natural
Bridge Road, St. Louis, Missouri 63121, U.S.A.; e-mail adoust@
umsl.edu.
1
Manuscript received November 2000; revised manuscript received March 2001.
697
INTERNATIONAL JOURNAL OF PLANT SCIENCES
698
Table 1
Numbers of Sepals, Petals, Stamens, and Carpels for Drimys
winteri Found in This Study and in the Studies
of Smith (1943) and Tucker (1959)
Flower organ
Sepal
Petal
Stamen
Carpel
This study
Smith (1943)
Tucker (1959)
2–3
6–14
16–42
4–12
2–3
6–14
24–35
4–10
2
7–9
20–30
5–7
shape of the apical meristem was responsible for irregularities
in the position of primordia. Erbar and Leins (1983) also noted
that the lateral flowers of D. winteri initiate two lateral sepals,
whereas the terminal flowers initiate the calyx as an annular
primordium encircling the meristem. All authors tried to explain floral organ arrangement as deviations from Fibonacci
phyllotaxis. This phyllotaxis is the most common pattern of
organ arrangement in plant systems (Jean 1994) and is characterized by a divergence angle between successively initiated
primordia of 137.5" and parastichy pairs (the opposing sets
of spirals of organs visible on plants) that are consecutive members of the Fibonacci series (1, 2, 3, 5, 8, 13, and 21, etc.).
Hiepko (1965, 1966) considered that some flowers of D. winteri showed deviations from Lucas phyllotaxis, an arrangement
of organs characterized by a divergence angle of 99.5" and
parastichy pairs that are consecutive members of the Lucas
series (1, 3, 4, 7, 11, 18, and 29, etc.). Doust (2000) compared
patterns of floral ontogeny and morphology amongst the different genera in Winteraceae, including Drimys, and established that the basic pattern of floral organ arrangement in the
family was one of pairs and whorls of organs, although spiral
organ arrangements were also observed. Variation in the number and arrangement of floral organs was not explored.
The studies noted above have established that flowers of D.
winteri exhibit large amounts of variation in organ number
and arrangement. However, there has not yet been an attempt
to understand this variation from a detailed developmental
perspective. The aim of the work presented here is to document
variation in organ number and arrangement in flowers of D.
winteri and to investigate, through scanning electron microscopy (SEM), how this variation arises. In addition, the high
levels of variation in flowers of D. winteri make it possible to
examine the relationship between floral meristem dynamics
and patterns of primordial initiation in a way that is not possible in morphologically invariant flowers. Therefore, the flowers of D. winteri constitute a natural experiment in variation,
and the study of their morphology and development provides
a complement to mutant-based approaches on plants such as
Arabidopsis and Antirrhinum (e.g., Clark et al. 1993; Crone
and Lord 1993; Rasmussen and Green 1993).
Buds and flowers of D. winteri were collected from five cultivated trees growing on the Parkville campus of the University
of Melbourne, with samples being taken over the flowering
season of the trees (July–October, 1995–1997). Additional material was collected from two trees cultivated at Pirianda Gardens, Olinda, in the Dandenong Ranges on the outskirts of
Melbourne, and from a tree cultivated on the campus of the
Victoria University of Wellington in New Zealand.
Inflorescence and Flower Development
Various terms have been used for describing reproductive
structures in D. winteri; here, the term inflorescence is the same
as the terms flowering branch (Tucker 1959) and inflorescence
(Vink 1970), whereas the term uniflorescence is the same as
the terms inflorescence (Tucker 1959) and florescence (Vink
1970). These terms are illustrated in figure 1.
Buds and flowers of all ages were fixed in formalin–acetic
acid–60% ethanol (10 : 5 : 85 v/v) for 3–5 d before being
transferred to 70% ethanol for storage. Voucher specimens are
stored in the herbarium at the University of Melbourne. Before
dissection, buds were dehydrated in an ethanol concentration
series and stained with 1% acid fuchsin in 95% ethanol to
enhance contrast when viewed under the dissecting micro-
Material and Methods
Collection of Materials
Drimys winteri var. chilensis (Gray) A.C. Smith is a small
tree or shrub native to central Chile (Smith 1943), a plant that
has been cultivated as an ornamental in temperate climes. It
flowers in southern Australia in late winter and early spring.
Fig. 1 Inflorescence and uniflorescence morphology
DOUST—FLORAL VARIATION IN DRIMYS WINTERI
Fig. 2 Position of initiation of the first petal primordium relative
to the major axis of the meristem. The smaller the angle (A) between
the primordium, major axis, and meristem center, the closer to the
pole of the elliptical meristem the primordium is situated.
scope. After dissection, buds were critical-point dried,
mounted on aluminum stubs, and sputter-coated with gold
before being imaged in either a Jeol 840 or a Phillips XL30
scanning electron microscope. The position of the microscope
stub was adjusted so that each specimen was viewed from
directly above (polar view) or from the side (lateral view).
Negatives of scanning electron microscope images were digitized at 600 dpi using a flatbed scanner, and digital images
were analyzed using the public domain National Institutes of
Health (NIH) Image program (developed at the U.S. NIH and
available on the Internet at http://rsb.info.nih.gov/nih-image/)
on a Macintosh PowerPC 5200/75. NIH Image was used to
measure flower and floral meristem area, perimeter, and the
length of the major and minor axes of the best-fitting ellipse
in polar view. The best-fitting ellipse is calculated by taking
the maximum width of the bud as the ellipse major axis and
determining the length of the minor axis in order to make the
ellipse area agree with the object area (Russ 1995). The shape
699
of the bud in polar view was measured as the aspect ratio of
the bud (flower aspect ratio), calculated by dividing the length
of the major axis of the best-fitting ellipse by the length of the
minor axis. The aspect ratio of a perfect circle will have a value
of 1, whereas increasingly narrow elliptic figures will have
increasingly larger values. Each bud was tracked through the
dissection and imaging procedures so that measurements on
the final image could be correlated with the bud’s original
position within the inflorescence.
To test whether there was a relationship between the shape
of the meristem (as measured by aspect ratio) and the position
of initiation of the first petal primordium, the angle between
the line of the major axis of the best-fitting ellipse and the line
connecting the primordium and the center of the meristem was
measured. When this angle is small, the primordium is close
to the major axis and thus close to one of the poles of the
ellipse of the meristem. When this angle is large, the primordium is being initiated on one of the sides of the meristem
away from the poles of the ellipse (fig. 2). Statistical analyses
were performed with SPSS (1995) and Minitab (1995).
Organ Arrangement
Analysis of organ arrangement was performed on buds in
mid-development with more than 20 organ primordia initiated. Mature flowers were not used because the arrangement
of the organs can be distorted by changes in the size and shape
of the flower just prior to and at anthesis.
The first step in characterizing floral organ arrangement was
to identify the parastichy pair, the pair of opposing sets of
spirals that are visible winding around the flower axis (denoted
as [m,n], with m spirals winding in one direction and n spirals
winding in the other) (fig. 3A). Where necessary, stereoscopic
pairs of images were used to determine the relative height of
primordia and to check the accuracy of the overlain spirals.
Fig. 3 Measuring phyllotactic patterns in Drimys winteri. A, Fibonacci pattern with five clockwise spirals and eight anticlockwise parastichy
spirals (5,8). The primordia have been numbered consecutively from the most recently initiated. B, Fibonacci pattern (5,8) showing the divergence
angle (d; ca. 137.5") between primordia 1 and 2. The ratio of the length of the radii of primordium 2 to primordium 1 is the plastochron ratio.
Scale: A, B, 100 mm.
700
INTERNATIONAL JOURNAL OF PLANT SCIENCES
Fig. 4 Simulation of the effect of divergence angle on variation in
plastochron ratio. Eleven primordia were placed sequentially around
the circumference of the ellipse in an anticlockwise direction according
to the Fibonacci (137.5") or Lucas (99.5") divergence angles (angle
between primordia 1 and 2; indicated by arrow). The distance of each
primordium to the center was used to calculate the plastochron ratio
between successive primordia.
More than one pair of spiral sets can occasionally be seen, but
the pair chosen was the one whose angle of intersection between opposing spirals was closest to 90" (as recommended
by Jean 1994).
Following identification of the parastichy spirals, primordia
were numbered in the presumed sequence of initiation using
the Bravais-Bravais theorem (Jean 1994), which holds that
primordia on one of a set of n parastichy spirals are each
separated on that spiral by n (e.g., where there are five spirals
in one parastichy set, consecutive primordia on one of those
spirals are separated by five, as in 1, 6, and 11, etc.). On some
buds the oldest primordia could not be numbered because they
were obscured by younger primordia. In all of the results presented, primordium 1 is the youngest and most recently
initiated.
Two measurements were used to analyze variation in floral
organ arrangement (fig. 3B). The first was divergence angle,
the angle made between two successively initiated primordia
and the center of the meristem. This quantity can be measured
for all pairs of sequentially initiated primordia in a flower.
Divergence angle has been considered an intrinsic property of
the flower (reviewed in Jean 1994), but in this study it is merely
used as a measure of the spatial arrangement of primordia.
The second measurement made was of plastochron ratio, calculated by dividing the distance to the center of the meristem
of one primordium by the distance to center of the next initiated primordium. If the floral meristem remains relatively
constant in size as it initiates organs, the plastochron ratio
becomes a measure of the relative rate of organ initiation.
Measurements of the distance to center of each primordium
were not used to compare relative rates of organ initiation
between flowers because of the confounding effect of the curvature of the meristem. Plastochron ratio is independent of the
curvature of the meristem because it uses the ratio of the distances from the center of the meristem of two sequentially
initiated primordia, which, given a more or less symmetrically
curved meristem, should be on areas of similar curvature of
the meristem.
Computer programs to calculate divergence angles and plastochron ratios were written in MATLAB (MathWorks 1992)
and are available from the author. The center of the floral
meristem was calculated by taking the mean of the X and Y
coordinates of the primordia. After calculating divergence angles, graphs of divergence angle versus primordium number
were produced. In many cases these showed periodic oscillations in divergence angle that were possibly the result of a
misalignment of the actual center of the meristem caused by
the fact that the meristem was not perfectly perpendicular to
the axis of the photograph (after Ryan et al. 1991). To correct
possible distortions in the images, the position of the center
was changed in a stepwise fashion relative to the positions of
the primordia until the center position was found where the
standard deviation of the divergence angles was at a minimum
(after Ryan et al. 1991). The new X and Y coordinates of the
primordia and the recalculated divergence angles were then
measured for further analysis. In multijugate systems (where
there is more than one spiral of initiating primordia), each
spiral was analyzed separately. For whorled systems the mean
and standard deviation of the divergence angles were calculated for primordia within each whorl.
Plastochron ratios were calculated for all pairs of primordia.
In whorled flowers, the plastochron ratio was calculated between the different whorls using a geometric mean formula
derived from Rutishauser (1998):
Rm p
! distance between primordia and center for whorl 2
! distance between primordia and center for whorl 1
[
where m is the number of primordia in a whorl.
1
]
m
,
Fig. 5 Uniflorescence development. A, Uniflorescence meristem initiating lateral floral meristems and floral bracts (bracts have been removed).
B, Uniflorescence meristem has transformed into the terminal floral meristem and is starting to initiate the circular calycine calyptra. Lateral
uniflorescence meristems are still narrowly elliptic in shape and have not yet started to initiate sepals. C, Terminal flower with circular calyptra
primordium. The basal lateral flower of the right-hand side (1) has just initiated two lateral sepals, but the basal flower on the left-hand side
(2) has not yet started sepal initiation. The two basal flowers are almost opposite each other. The subtending bract for this uniflorescence would
be toward the base of this figure. The two lateral floral buds that are perpendicularly inserted relative to the uniflorescence bract (buds 1 and
2) have more rounded meristems than the floral meristems that are inserted parallel to the uniflorescence bract (bud 3). D, The lateral flower
on the right-hand side of the photo has initiated a secondary flower and subtending bracteole on its axis. The lateral flower has initiated the
calycine calyptra in a more or less circular ring, more similar to that of the terminal flower than to that of the lateral flower on the left-hand
side of the flower. E, Uniflorescence close to maturity, showing the rounded terminal floral bud and the spirally arranged elliptic lateral floral
buds. The free tips of the lateral buds are clearly visible. F, Lateral flower with its floral bract and a bracteole on the pedicel of the flower.
UM p uniflorescence meristem, LF p lateral floral meristem, FB p floral bract, TF p terminal floral meristem, Ca p calycine calyptra, ls p
lateral sepals, 2F p secondary flower, bt p bracteole. Scale: A, B, 200 mm; C, 100 mm; D, 200 mm; E, F, 1 mm.
INTERNATIONAL JOURNAL OF PLANT SCIENCES
702
Fig. 6 Floral meristem shape (aspect ratio in polar view) versus
stage of flower development. The stages of flower development represent developmental points in a linear sequence, but the difference
between sequential pairs of stages does not represent equal time
increments.
Means and standard deviations of divergence angles and
plastochron ratios were calculated for each flower. Means for
each phyllotactic pattern were calculated from the means for
divergence angle and plastochron ratio from each flower exhibiting that particular pattern.
The relationship of meristem shape to divergence angle and
plastochron ratio was examined by regression techniques to
determine whether an elliptic meristem influences variation in
these parameters. Variation was expressed as coefficients of
variation (CV), because these are independent of the magnitude
of the means and thus allow comparison of variation across
different means. The coefficients are calculated as follows (Sokal and Rohlf 1995):
coefficient of variation p
standard deviation # 100
.
mean
To further investigate the relationship between plastochron
ratio, divergence angle, and the shape of the floral meristem,
a simulation was conducted of the effect of two different divergence angles on plastochron ratio for an identical elliptic
meristem. Two sets of primordia were used, one arranged according to the divergence angle of a Fibonacci phyllotactic
pattern (d p 137.5") and the other according to that of a Lucas
phyllotactic pattern (d p 99.5") (fig. 4). Coefficients of variation of the plastochron ratios were compared between the
two divergence angle patterns.
of their parts. For each uniflorescence, the tree from which it
came, the position of the flowers, and how many petals, stamens, and carpels each flower contained were recorded. The
only missing case was from tree 5, where only four examples
of two-flowered uniflorescences could be found. The missing
value for this case was estimated from the mean of the values
of the other four cases in that cell of the analysis to avoid the
added complexities of an unbalanced design.
A MANOVA of the variation in petal, stamen, and carpel
number was performed. Differences in organ number were
compared between trees (tree) and between the different positions of the flowers on the uniflorescence axis (flower position). It was not possible to include uniflorescence class as a
variable in the same analysis as tree and flower position because the number of flower positions varied between uniflorescence classes. This would have resulted in different parts
of the analysis containing different numbers of levels of the
variables, a situation unable to be analyzed using standard
MANOVA techniques (Winer et al. 1991). For this reason the
effects of differences between trees and between flower positions were analyzed separately for each uniflorescence class.
The design of the MANOVAs was an unreplicated split-plot
design (Winer et al. 1991) with the between-plot analysis examining the variation between trees and the within-plot analysis examining the variation between flower positions and the
interaction between tree and flower position (tree # flower
position). In this design the plots are the randomly selected
branches (branchs) bearing the uniflorescences of each uniflorescence class.
The MANOVAs tested for multivariate differences between
the factors of tree, flower position, and tree # flower position
using Pillai’s trace test (Norusis 1993; Tabachnick and Fidell
1996). Each MANOVA was followed by univariate or, where
significant correlations between petal, stamen, and carpel values were found, step-down analyses to determine which variables were contributing to significant multivariate test results.
The order of variables used for the step-down analyses was
petal, stamen, and carpel, which follows the order of initiation
of the floral organs. Because the univariate tests and step-down
analyses for each MANOVA involved three ANOVAs, the sig-
Organ Number
To analyze variation in organ number, five samples of each
of two-, three-, four-, five-, six-, and seven-flowered uniflorescences (referred to as uniflorescence classes 2–7) were collected in 1995 from each of the five trees growing on the
campus of the University of Melbourne. Each uniflorescence
was collected from a randomly selected branch. Uniflorescences were harvested when a few of the flowers had opened
but before any of the flowers had aged sufficiently to lose any
Fig. 7 Positional relationships between uniflorescence and floral
bracts and the shape of the floral meristem in flowers in different
positions within the uniflorescence. Arrows indicate direction of pressure exerted by bracts: A–D, pressure of individual floral bracts; UB,
pressure of uniflorescence bract. Lateral primordia are numbered in
order of initiation, as in fig. 5C.
DOUST—FLORAL VARIATION IN DRIMYS WINTERI
703
To check the reproducibility of the results obtained, they
were analyzed in conjunction with two other data sets. One
of these data sets comprised 31 uniflorescences collected in
1996 (the year following the main experiment) from tree 2 of
the five trees at the University of Melbourne. The other was
of 19 uniflorescences from a tree growing on the campus of
Victoria University of Wellington, New Zealand, collected in
1996. To compare the three data sets, the difference in numbers
of floral organs between the terminal flower and the average
of the lateral flowers for each uniflorescence was examined.
The entire data set for the New Zealand collections was used,
but 19 uniflorescences from each of the other data sets were
randomly chosen so that each data set was represented by an
equal number of uniflorescences.
Results
Fig. 8 Circumference of floral meristem versus stage of flower development. Overall, the mean circumference of terminal meristems is
significantly larger than that of lateral meristems.
nificance level for each ANOVA was adjusted so that the overall significance level (experiment-wide level) did not exceed
P ! 0.05. The Bonferroni adjustment was used, making the
appropriate adjusted level for each individual test P !
0.0167 (Tabachnick and Fidell 1996). All assumptions of univariate and multivariate analyses were satisfied except for
moderate deviations from the assumption of sphericity. These
deviations were accounted for by decreasing the degrees of
freedom for the F-test using the Greenhouse-Geisser corrections obtained through reconfiguring the analyses to run as
repeated measures designs (Winer et al. 1991; Norusis 1993).
No transformation of the data was found to be necessary.
Approximate variance components were calculated to estimate the amount of variation in petal, stamen, and carpel
numbers explained by differences between the factors tree,
flower position, or tree # flower position. To do this, the
variance associated with each of these factors, over and above
the residual variance, was calculated and expressed as a percentage of the total variation. Variance components are considered approximate because there is a mixture of fixed (tree
and flower position) and random (branch) factors present
(Brown and Mosteller 1991). To calculate the variance components it was necessary to arbitrarily set the term for random
error to zero, because there was no replication of flower positions within each uniflorescence (there is only one flower in
each flower position within a uniflorescence). However, random error must still play a role in the variation seen, but it is
confounded with the interaction between branch and flower
position. Values for individual variance components can only
be strictly compared within each uniflorescence class, although
patterns of variation in the components can be compared
across uniflorescence classes.
Post hoc comparisons using the Scheffé method (Sokal and
Rohlf 1995) were performed on those factors that showed
overall significant effects. The significance level used for Scheffé
comparisons was the same as that used for the overall F-test;
thus, P ! 0.0167 (experiment-wide level of P ! 0.05) was used.
Uniflorescence and Floral Development
The shoot meristem in Drimys winteri produces leaf primordia during the vegetative phase and uniflorescence primordia during the production of the inflorescence. Both sets
of primordia are arranged spirally and, after the production
of the inflorescence, the shoot meristem reverts to producing
leaves.
Each uniflorescence initiates lateral floral bracts and flower
primordia in an acropetal direction (fig. 5A), before the uniflorescence meristem eventually transforms into the terminal
floral primordium (fig. 5B). One to eight lateral floral primordia are initiated, but they do not continue development
until the terminal floral primordium has commenced sepal initiation (fig. 5C). At this stage the lateral floral primordia have
a narrowly elliptic and falcate shape. The first flowers of the
uniflorescence occur in lateral positions (at right angles to the
subtending bract) and are subopposite to each other (fig. 5B).
These two lateral buds may or may not be at similar stages
of development. Subsequent lateral flowers are initiated in a
more or less spiral pattern (fig. 5C–5E). At all stages, lateral
flowers usually appear more elliptic in transverse section than
do terminal flowers (fig. 5D, 5E). On rare occasions, lateral
floral primordia produce secondary lateral floral primordia
together with their subtending bracts (bracteoles), indicating
that a secondary lateral branch in these uniflorescences is possible (fig. 5D, 5F).
Floral meristem shape is variable.
The cross-sectional
shape of the floral meristem in terminal flowers is usually more
or less circular throughout development (aspect ratio close to
unity), whereas the floral meristem in lateral flowers is more
elliptic, especially during early development (aspect ratio between 1.5 : 1 and 2.5 : 1) (fig. 6). Lateral flowers in different
positions within the uniflorescence show different amounts of
deviation from circularity, with buds in positions 1 and 2 (lateral to the subtending bract of the uniflorescence) being more
rounded in shape than those in positions 3 and 4 (more or
less parallel to the subtending bract of the uniflorescence) (fig.
5C, 5D). The shape of the lateral floral meristems appears
related to the position of the flower in the uniflorescence, possibly because of the additive or opposing pressures of the uniflorescence and floral bracts (figs. 5C, 7). There is a large
704
INTERNATIONAL JOURNAL OF PLANT SCIENCES
Fig. 9 Calyx and corolla development. A, Lateral flower showing the two lateral sepals initiated at each pole of the elliptic meristem; floral
bract situated toward base of micrograph. B, Lateral flower showing the growth of the calyptra starting to lift the sepals up above the floral
meristem. C, Lateral flower bud; the calyptra bearing the lateral sepals at its tip has almost entirely enclosed the floral meristem. D, Lateral
flower that has developed an uneven ring of calyx rather than two sepal tips. E, Polar view of the large, domed floral meristem at the start of
initiation of the petals. F, Side view of the floral meristem at the start of initiation of the petals. G, Terminal flower with petals initiated all
around circumference of floral meristem. H, Lateral flower in side view with a petal initiated at one of the poles of the elliptic meristem. I,
Lateral flower with elliptic floral meristem; petals have been initiated at the lateral poles of the meristem. M p meristem, Ca p calyptra, ls p
lateral sepal, p p petal. Scale: A, B, 100 mm; C, 200 mm; D–I, 100 mm.
amount of variation in cross-sectional shape in both terminal
and lateral flowers at all stages of development.
Terminal flowers have floral meristems with significantly
larger circumferences than lateral flowers at all stages of development, although both display considerable variation (fig.
8). No significant differences were observed in the height of
the meristem between terminal and lateral flowers.
Floral development varies between flowers in a uniflorescence. In terminal flowers the calyx is initiated as an annular
ring of tissue around the more or less circular floral meristem
when the meristem is ca. 120 mm high and 180 mm in diameter
(fig. 5B). In lateral flowers two sepal tips are initiated when
the floral meristem is ca. 130 mm in height, 190 mm in breadth
(from one lateral side to the other), and 100 mm in depth (from
abaxial to adaxial sides). Each sepal tip is initiated laterally
with respect to the subtending floral bract (fig. 9A), but most
growth of the calyx that produces the calyptra of the bud
occurs in a region encircling the meristem and beneath the
sepal tips (fig. 9B). This growth subsequently bears the sepal
tips aloft. The calyptra eventually encloses the developing
flower (fig. 9C), but at no stage is there any fusion of the tips
or the leading edge of the calyptra in either terminal or lateral
flowers. In some lateral flowers the calyptra does not develop
free sepal tips but grows throughout as an annular ring (fig.
9D), as in terminal flowers. At anthesis the calyptra splits into
two or three segments and the petals unfold. The reflexed
calyptra segments eventually wither and abscise.
There is a delay between calyx and petal initiation. In
both terminal and lateral flowers there is a considerable delay
between initiation of the calycine calyptra and the corolla;
petal primordia appear only when the calyptra has almost
enclosed the floral meristem. At this stage the floral meristem
is large and domed and petals initiate low down on the flank
of the dome (fig. 9E, 9F). When first visible the petal primordia
are ca. 1000 mm2 in size, and the apical meristem is ca. 100
mm in height, 200 mm in breadth, and 150 mm in depth. When
petal primordia are first being initiated, the circumference of
the apical meristem varies significantly more than the arc of
the circumference occupied by each petal primordium (coef-
DOUST—FLORAL VARIATION IN DRIMYS WINTERI
Fig. 10 Magnitude of angle between primordium and major axis
of flower versus flower aspect ratio, showing that the angle decreases
as the meristem becomes more elliptic (i.e., aspect ratio increases)
(P ! 0.01, r 2 p 12.6).
ficient of variation for floral meristem p 14.6%, primordium
arc p 4.9%, difference significant at P ! 0.01).
Position of initiation of first few petal primordia depends
on flower shape. In flowers with meristems that are gently
elliptic to more or less circular (terminal and some lateral flowers), there is no discernible preference in the position of initiation of the petal primordia. In these flowers subsequent primordia are quickly initiated all around the circumference of
the meristem (fig. 9G) and are either in a spiral or whorled
pattern. However, in lateral flowers whose meristems are more
elliptic in shape, the first petals to be initiated generally arise
laterally on the meristem, in the same region as the free sepal
tips (fig. 9H, 9I). This trend is significant (fig. 10). Further
primordia are initiated at the poles of the elliptic meristem
before primordia start to be initiated around the remainder of
the circumference. Regardless of where the first formed petal
primordia initiate in terminal and lateral flowers, the various
phyllotactic patterns are not related to the phyllotaxy of the
calyx, whether decussate in lateral flowers or nonexistent in
terminal flowers (where the calyx is initiated as a ring of tissue).
The delay between the initiation of the calyx and petals apparently allows the petals to start a new phyllotactic pattern.
The phyllotactic pattern of petals is propagated throughout
stamens and carpels. Initiation of stamens follows the petals
without delay. Stamens are initiated on the almost vertical
flanks of the floral meristem and are first visible at a size of
ca. 700 mm2, at a stage when the meristem is large and domed
and is ca. 90 mm in height, 200 mm in breadth, and 160 mm
in depth (fig. 11A, 11B). Several tiers of stamens are initiated,
filling the flanks of the floral meristem (fig. 11C, 11D). In
lateral flowers that are narrowly elliptic in outline the stamens
tend to be initiated first toward the poles of the meristem (fig.
11E), although stamen primordia completely surround the
meristem by the end of the period of stamen initiation (fig.
11F). When viewed from the side, terminal flowers often have
at least three tiers of stamens (fig. 11C), whereas lateral flowers
usually only have two.
Initiation of carpels follows the stamens without delay, and
as the growth of the floral meristem ceases, the carpels eventually develop all over the flattened apex of the dome (fig.
11G–11K). However, occasionally a small area of the floral
705
meristem remains uncommitted (fig. 11H–11L). Carpels initiate as circular primordia and are first clearly apparent when
their area is ca. 3000 mm2, and the floral meristem has a
breadth and depth of ca. 190 mm and 150 mm, respectively,
and height of ca. 20 mm (fig. 11G). As the carpels develop the
adaxial face becomes depressed (fig. 11H, 11I) and forms a
slit (fig. 11J, 11K). At anthesis the carpels are oblong to obovoid with a distal and adaxial stigmatic crest (fig. 11L).
During development of the inflorescence the uniflorescence
bracts tightly enclose the developing uniflorescences and they
only spread open when the floral buds are relatively well developed (fig. 12A). Elongation of the uniflorescence peduncles
starts with the more proximal uniflorescences, carrying the
flowers away from the uniflorescence bracts (fig. 12B). Soon
after, elongation of the pedicel of each flower carries the flower
away from the floral bract and the flower opens, with the
flowers on the uniflorescences toward the base of the inflorescence opening before those toward the apex (fig. 12C). Basal
uniflorescences also have more flowers than more distal ones
(fig. 12C). Fully open flowers are actinomorphic, with the petals held more or less horizontally around the dome of the
receptacle (fig. 12D). All of the organs are free, although occasionally two adjacent primordia fuse together in early development and produce a double organ at maturity.
Arrangement of Petals, Stamens, and Carpels
Many phyllotactic patterns occur. Floral phyllotaxis was
examined in 52 buds that were developmentally between midstamen and late carpel initiation. There were 14 buds with
whorled phyllotaxis (27%) distributed in three patterns (5,5),
(6,6), and (7,7). The rest of the buds had spiral phyllotaxis,
of which 11% showed Fibonacci phyllotaxis (divergence angle
close to 137.5" and parastichy pairs of [3,5] or [5,8]), 26%
showed Lucas phyllotaxis (divergence angle close to 99.5" and
parastichy pairs of [4,7] or [7,11]), 8% were multijugate, and
55% showed some other form of phyllotaxis (figs. 3, 13, 14;
table 2). All buds examined could be described by a recognizable pattern, although in one case it appeared that an extra
primordium had initiated that did not fit onto the parastichy
spirals (fig. 13E, asterisk). Mean divergence angles for each
phyllotactic pattern are close to the limit divergence angles for
these patterns given by Jean (1994) (table 3). Fewer terminal
buds were examined than lateral buds, and they had fewer
patterns (table 4). No major preference in the direction of the
spiral of sequentially initiated primordia was evident; 21 of
the flowers had primordia arranged in a clockwise spiral, and
17 had primordia arranged in a counterclockwise spiral.
Basal petal primordia are the start of both sets of parastichy
spirals, and the number of primordia in the first tier corresponds to the first term of the parastichy pair (e.g., the parastichy pair [3,5] has three petal primordia in the first tier of
initiated petals). In terminal flowers there was no significant
relationship between the shape of the floral meristem and the
number of parastichy spirals (signified by the first term of the
parastichy pair). However, in lateral flowers there was a significant positive relationship, so that flowers with more elliptic
meristems tended to have phyllotactic patterns with more parastichy spirals (fig. 15).
Variation in phyllotactic pattern is related to the shape of
706
INTERNATIONAL JOURNAL OF PLANT SCIENCES
Fig. 11 Androecium and gynoecium development. A, B, Terminal flower; stamens are being initiated on the almost vertical sides of the large,
domed floral meristem. A, Side view. B, Polar view. C, D, Terminal flower; three tiers of stamens have initiated, filling up the sides of the floral
meristem. C, Side view. D, Polar view. E, Lateral flower, polar view, showing stamens initiated at the poles of the elliptic meristem. F, Lateral
flower, polar view, showing stamens initiated all around the circumference of the floral meristem. G, Lateral flower, polar view, showing carpels
developing on an almost flat surface of floral meristem. H, Lateral flower, polar view; the ventral invagination in each carpel is starting to appear.
I, Terminal flower, side view, showing the carpels as the last tier of organs initiated by the floral meristem. J, K, Lateral flower, polar view. J,
The ventral invagination in the carpels is becoming a slit. Note the uncommitted floral meristem in the center of the flower. K, Slits in carpels
are almost closed, with no visible uncommitted floral meristem. L, Carpels in center of flower at anthesis showing opening of slit now closed
by papillae to form the stigmatic crest. The uncommitted floral meristem is visible in the center of the flower. M p floral meristem, p p petal,
s p stamen, c p carpel. Scale: A, 100 mm; B, 200 mm; C, D, 100 mm; E, 50 mm; F–K, 100 mm; L, 2 mm.
the meristem. Periodic variation in divergence angle and
plastochron ratio was found within individual flowers, even
after adjusting the center of the meristem to give the least
variability in measured divergence angles (fig. 16A–16F). Fluctuations occurred in both whorled and spiral phyllotactic patterns, but the size of the fluctuations was not constant across
flowers (e.g., cf. fig. 16A, 16E). However, linear regressions
of plastochron ratio versus age of the primordia showed no
significant relationship, indicating that plastochron ratio does
not change during the initiation of the petals, stamens, and
carpels.
There was a significant relationship (P ! 0.001) between aspect ratio of the flower and variation in divergence angle (fig.
17), with flowers whose meristems were narrowly elliptic being
the most variable in divergence angle. The amount of variation
of the coefficients of variation for divergence angle explained
DOUST—FLORAL VARIATION IN DRIMYS WINTERI
707
Fig. 12 Inflorescence and flower morphology. A, Developing inflorescence; uniflorescence bracts open as the uniflorescences expand. B,
Peduncles on the uniflorescences toward the base of the inflorescence have elongated, and the flowers they bear are about to open. C, Inflorescence
with basal uniflorescences at anthesis and distal uniflorescences just prior to anthesis. D, Lateral view of flower showing more or less horizontally
held petals, whereas stamens and carpels are arranged on the dome of the floral receptacle. U p uniflorescence, UB p uniflorescence bract.
Scale: A, 12 mm; B, 16 mm; C, 30 mm; D, 4 mm.
by the elliptic shape of the meristem was 22%. There was no
significant relationship between variation in plastochron ratio
and the elliptic shape of the floral meristem (fig. 17). However,
the results of the simulation of variation in plastochron ratio
by sequentially positioning primordia around an elliptic meristem at either the Lucas divergence angle (99.5") or the Fibonacci divergence angle (137.5") showed that primordia positioned at 99.5" had 50% more variation in plastochron ratio
than primordia positioned at 137.5", even though the mean
plastochron ratio was very similar (table 5). Thus, the amount
of variation in plastochron ratio depends on both the shape
of the meristem and the size of the divergence angle (which is
a function of the phyllotactic pattern).
In most cases the spiral of sequentially initiated primordia
consisted of an orderly sequence of petals, stamens, and carpels. However, in some flowers there was a discrepancy between the position of some of the organs in the sequence of
initiation and their identity. In figure 13D, primordia 1–7 are
carpels, primordia 8 and 9 are stamens, and primordium 10
is a carpel. Primordia after primordia 10 are stamens. This
indicates that the sequence of initiation inferred from the phyl-
lotactic pattern may not always correspond to the actual sequence of differentiation of organs in a flower.
Variation in Organ Number
Organ number significantly differs between flowers in different parts of the uniflorescence. Petal numbers ranged between 6 and 14, stamen numbers between 16 and 42, carpel
numbers between 4 and 12, and total numbers of floral organs
between 30 and 60 (table 1). The distribution of the variation
was analyzed by a MANOVA test (Pillai’s trace statistic), which
showed that in all of the six uniflorescence classes there were
significant differences in organ number between flower positions
on the uniflorescence (table 6). In two of the six classes (classes
5 and 7) there were small but significant differences between
trees in addition to that difference found between flower positions. The nonsignificant interaction tree # flower position indicates that the pattern of differences between flower positions
was similar in all trees.
The pattern of variation in petal, stamen, and carpel numbers was analyzed separately in univariate analyses and step-
Fig. 13 Phyllotactic patterns in Drimys winteri. A, 3(2,3) (d p 45.84") trijugate Fibonacci pattern. B, 2(3,4) (d p 49.75") bijugate Lucas
pattern. C, 4,5 (d p 77.96"). D, 4,7 (d p 99.5") Lucas pattern. Note that primordia 1–7 are carpels, primordia 8 and 9 are stamens, and
primordium 10 is a carpel. E, 5,5 (d p 72"). Note that one stamen primordium (*) does not fit onto the parastichy spirals. F, 5,6 (d p 64.08").
Scale: A–F, 100 mm.
708
Fig. 14 Phyllotactic patterns in Drimys winteri. A, 5,7 (d p 151.14"). B, 6,6 (d p 60"). C, 6,7 (d p 54.4"). D, 7,7 (d p 51.43"). E, 7,8
(d p 47.25"). F, 8,9 (d p 41.64"). Scale: A–F, 100 mm.
709
INTERNATIONAL JOURNAL OF PLANT SCIENCES
710
Table 2
Phyllotactic Patterns Grouped According to Type
Phyllotactic
pattern
Percentage of
all buds
Frequency
Whorled
Spiral:
Fibonacci
Lucas
Multijugate
Fibonacci-type series
(other than Lucas)
Total
Percentage of
spiral buds
Predicted
percentage
(Jean 1994)
14
26.9
…
…
4
10
3
7.7
19.2
5.8
10.5
26.3
7.9
92
2
5
21
40.4
55.3
1
52
100
down analyses (step-down analyses only applied when significant correlations were found between petal, stamen, and
carpel numbers using Bartlett’s test). In all cases the interaction
tree # flower position was not significant. The results can be
summarized as follows (table 7). (1) Petals: Petal numbers
showed significant differences between trees only for uniflorescence class 7, with 17% of the variance in petal number
explained by differences between trees. Uniflorescence classes
2, 5, 6, and 7 showed significant differences between flower
positions, with these differences accounting for 10%–20% of
the variance in petal number. (2) Stamens: Stamen numbers
were significantly different between flower positions for all
uniflorescence classes, with 36%–51% of the variance in stamen numbers explained by differences between flower positions. There were no significant differences between trees. (3)
Carpels: Carpel numbers were significantly different between
trees for uniflorescence classes 5 and 7, with 19% and 25%
of the variance in carpel numbers explained by these differences. There were significant differences between flower positions for uniflorescence classes 2, 5, 6, and 7, with these
differences accounting for 3%–13% of the variance in carpel
number.
The difference between flower positions accounts for much
100
100
of the total variation in floral organ number, particularly in
stamen number (36%–51%). Significance levels vary considerably between uniflorescence classes for differences in petal
and carpel numbers but not for differences in stamen numbers.
Differences in organ number are primarily the result of stamen
number differences between terminal and lateral flowers.
The results of the post hoc multiple comparison tests are summarized over all uniflorescence classes as follows (table 8):
1. Pairwise comparisons between trees: petal and stamen
pairwise comparisons between trees were not significant
in any case. Carpel comparisons were significant in 5%
of the comparisons.
2. Pairwise comparisons between lateral flowers in a uniflorescence: petal showed significant differences between
pairwise comparisons of lateral flowers in 9% of cases.
Stamen and carpel comparisons were not significant.
3. Pairwise comparisons between the terminal and each of
the lateral flowers in a uniflorescence: stamen showed
significant differences between pairwise comparisons of
terminal (flower position 1) and lateral flowers (flower
positions 2–7) in all cases. Petal and carpel comparisons
were significant in 14% and 19% of cases, respectively.
4. Comparisons between the terminal and an average of the
Table 3
Type and Frequency of Phyllotactic Patterns Found in Drimys winteri
Phyllotaxis
3(2,3) trijugate Fibonacci
(3,5), (5,8) Fibonacci
2(3,4) bijugate Lucas
(4,5)
(4,7), (7,11) Lucas
(5,5)
(5,6), (6,11)
(5,7)
(6,6)
(6,7), (7,13)
(7,7)
(7,8)
(8,9)
Frequency
Theoretically
correct angle
1
4
2
1
10
1
6
2
7
9
6
2
1
45.84
137.51
49.75
77.96
99.50
72.00†
64.08
151.14
60.00†
54.40
51.43†
47.25
Not given
Mean divergence
angle (SE)
46.47
137.52
50.55
76.56
99.36
70.94
65.01
150.92
59.95
54.95
50.04
47.93
41.64
(0.33)
(0.34)
(0.17)
(0.24)
(0.14)
(0.56)
(0.15)
(0.45)
(0)
Mean plastochron
ratio (SE)
1.09
1.08 (0.001)
1.1 (0.03)
1.09
1.07 (0.005)
1.07
1.07 (0.013)
1.07 (0.018)
1.05 (0.003)
1.07 (0.008)
1.04 (0.001)
1.05 (0.016)
1.07
Note. Theoretically correct angle and mean divergence angle are measured in degrees. The limit divergence angle for each
pattern is given in column three (from Jean 1994, p. 39), except for whorled patterns, where the values (†) are those given by
dividing 360" by the number of primordia in each whorl.
DOUST—FLORAL VARIATION IN DRIMYS WINTERI
711
Table 4
Distribution of Phyllotactic Patterns between Terminal and Lateral Flowers
Position of bud
Terminal
Lateral
Number of buds
Phyllotactic patterns
15
37
3(2,3), (4,7), (5,7), (6,6), (7,7), (7,8)
(3,5), (5,8), 2(3,4), (4,5), (4,7), (7,11), (5,5), (5,6), (6,11), (6,6), (6,7), (7,13), (7,7), (7,8), (8,9)
lateral flowers in a uniflorescence: stamen showed significant differences between comparisons of terminal
(flower position 1) and averaged lateral flowers in all
cases. Petal and carpel comparisons were significant in
17% and 50% of cases, respectively.
These comparisons show that, in general, the difference between terminal and lateral flowers explains most of the total
variation in organ number, with terminal flowers having a
mean of 7.45 more floral organs than lateral flowers (table 9).
The increase in the number of stamens between lateral and
terminal flowers in a uniflorescence (22.1%) is significantly
greater (P ! 0.001) than the increase in petal or carpel number
(5.8% and 9.2%, respectively; table 9).
Although most variation in organ number is between terminal and lateral flowers, some differences were also seen between the different lateral flower positions. In particular, the
two most basal lateral flowers had, on average, fewer of each
organ type than the more distal lateral flowers (data not
shown). These differences were not significant in the post hoc
pairwise comparisons between lateral flower positions using
the Scheffé test. However, this test is very conservative (Sokal
and Rohlf 1995), especially when used in conjunction with
the Bonferroni adjustment to the univariate P values, and a
larger data set would be needed to examine this phenomenon
statistically.
Comparison with other data sets confirms main findings.
Two other data sets were examined to see if they would show
results consistent with those reported above. No significant
differences were discovered among the three data sets in the
size of the difference in floral organ numbers between terminal
and lateral flowers (table 10).
Shape of the Floral Meristem
Before initiation of the calyx there is a major difference
between the narrowly elliptic shape of most lateral floral meristems and the more or less circular shape of the terminal floral
meristem. Each lateral floral meristem is initiated in the axil
between its subtending floral bract and the uniflorescence axis,
whereas the terminal meristem (which is the transformed uniflorescence meristem) terminates the shoot and lacks a subtending floral bract. After the lateral sepals are initiated the
floral meristems of lateral flowers are more circular in outline
than before, but they are still more elliptic than the floral
meristems of terminal flowers. Lateral floral meristems can
become more circular as they develop because the buds expand
as they grow out from the confines of the acute angle between
the subtending floral bract and the uniflorescence axis.
The shape of the floral meristems of lateral flowers is affected
both by the constraint of the subtending floral bract as well
as that of the bract subtending the entire uniflorescence. Floral
buds that are lateral to the subtending uniflorescence bract
have the pressure of the uniflorescence bract perpendicular to
the constraint of the floral bract, so that the two forces partially
counteract each other (fig. 7). However, floral buds in the same
plane as the uniflorescence bract have the pressure of the uniflorescence bract and that of the floral bract working in tandem
to increase the constraint on the floral meristem. This may be
particularly important for the basal flowers toward the outer
edge of the uniflorescence than for the more distal flowers
toward the center of the uniflorescence. Differences in shape
that may correspond to these pressures can be seen in the SEM
micrographs (fig. 5C, 5D), where meristems on the left- and
right-hand sides of the figures are lateral with respect to the
Discussion
Much of the variation in floral morphology of Drimys winteri is attributable to differences between terminal and lateral
flowers. In particular, terminal floral meristems are generally
larger, more circular in shape, and initiate more organs than
lateral floral meristems. Differences in size are linked to organ
number differences, whereas meristem shape appears particularly important in determining variation in organ arrangement. The developmental pathway of each flower depends on
subtle variations in size and shape of the floral meristem, resulting in a large degree of morphological variation. A number
of key events during development allow different floral morphologies to arise, of which the most important is variation
in the shape of the floral meristem from the very beginning of
floral initiation.
Fig. 15 Phyllotaxis versus flower shape. Phyllotactic pattern is represented as the first term of the parastichy spiral set. Regression of
phyllotaxis versus flower aspect ratio for terminal flowers was not
significant; regression of the same variables for lateral flowers was
significant (P ! 0.05, r 2 p 11.2).
712
INTERNATIONAL JOURNAL OF PLANT SCIENCES
terminal floral meristem must remain actively initiating primordia that will become stamens for a longer period of time
than do lateral meristems. Examination of mature buds of
terminal and lateral flowers indicates that there is generally an
extra tier of stamens in terminal flowers, indicating that the
floral receptacle is correspondingly longer in terminal than in
lateral flowers.
Variation in number of organs in the flowers of D. winteri
has been observed by previous workers, although the correlates
of that variation had not been elucidated. Smith (1943) remarked on the variable numbers of floral organs in different
species of Drimys, and he gave a range of variation for D.
winteri var. chilensis that is similar to that of this study (table
1). Tucker (1959) gave approximate limits for the variation in
floral organ number of the three trees that she examined, but
this variation was less than that found by either Smith (1943)
or that found in this study.
Endress (1987) has drawn attention to the fact that flowers
with spiral phyllotaxis are relatively plastic, so that organ number may vary depending on the position of the flower in the
branching system. Schöffel (1932) demonstrated that the number of organs in terminal flowers of Ranunculus (Ranunculaceae) may be greater than that in lateral flowers, a situation
similar to that found here. Endress (1978) showed that flowers
of members of the Hamamelidoideae could vary in number of
floral organs and that this variation was primarily attributable
to variation in stamen numbers.
Fig. 16 Divergence angle and plastochron ratio versus primordium
number for three flowers (flower 1 p A + B, 2 p C + D, 3 p E +
F). Regressions (dashed line of best fit) shown for plastochron ratio
are not significant. Lines connecting consecutive points are not continuous in C and D because the organs are arranged in whorls; only
the angles and plastochron ratios between primordia within each
whorl have been shown. W1 p whorl 1, etc.; m,n p parastichy pair;
l.d.a. p limit divergence angle for the phyllotactic pattern.
uniflorescence bract and are more rounded than the narrowly
elliptic meristems at the top and bottom of the figures (which
are in the same plane as the uniflorescence bract). Such differences in shape of the buds in the various positions within
the uniflorescence were first observed by Tucker (1959), who
noted that the lowest pair of lateral buds was relatively circular
in outline compared with the narrowly elliptic shape of the
next highest pair and that further buds became more circular
in outline again.
Initiation of the Calyx and the Petals
The lateral floral meristem produces two lateral sepals,
whereas the terminal meristem produces an annular calyx primordium, as first reported by Erbar and Leins (1983). The
two modes of initiation correlate with the shape of the meristem, as lateral meristems are much more elliptic than terminal
meristems at this stage. Occasional lateral flowers produce a
more or less complete annular calyx primordium, presumably
because of a more circular floral meristem. In lateral flowers
the sepals are uplifted by the growth of the continuous ring
of calycine tissue beneath them.
Floral Organ Number
The difference in size between terminal and lateral floral
meristems is correlated with a significant difference in number
of floral organs; terminal flowers have approximately seven
more organs than lateral flowers. Differences in stamen number appear to be primarily responsible for the overall difference, with terminal flowers having six to seven more stamens
than lateral flowers. The difference in stamen numbers is significant over all uniflorescence classes and across three different
data sets. The disproportionate increase in stamen, as opposed
to petal or carpel number (increase of approximately six stamens as opposed to 0.6 petals and carpels), indicates that the
Fig. 17 Variation in divergence angle and plastochron ratio versus
flower shape. Regression for coefficient of variation (CV) for divergence angle versus aspect ratio is significant (P ! 0.001, r 2 p 0.22);
regression for CV plastochron ratio versus aspect ratio is not significant
(r 2 p 0.02).
DOUST—FLORAL VARIATION IN DRIMYS WINTERI
Table 5
Plastochron Ratios Calculated for Each Pair of Primordia
in the Simulation of the Fibonacci and Lucas
Placement of Primordia in Figure 4
Primordia
pairs
1&2
2&3
3&4
4&5
5&6
6&7
7&8
8&9
9 & 10
10 & 11
Mean (SD)
Coefficient of
variation (%)
Fibonacci angle (137.5"):
plastochron ratios
Lucas angle (99.5"):
plastochron ratios
1.14
1.30
0.83
0.81
1.10
1.31
0.92
0.78
0.99
1.33
1.05 (0.21)
1.51
0.69
1.26
0.86
1.05
1.09
0.81
1.42
0.68
1.47
1.01 (0.32)
20
31.68
The most important developmental event allowing the diverse range of phyllotactic patterns to arise is the delay between
the initiation of the sepal and petal primordia. Because of this
delay, petal primordia do not have any positional cues from
the already initiated sepal primordia, and therefore the position in which they initiate is only affected by the overall shape
of the meristem. The effect of meristem shape on petal primordium position is similar to its effect on sepal position, with
primordia on narrowly elliptic meristems being initiated toward the poles of the meristem and primordia on circular
meristems being initiated at any point around the circumference. This pattern could be the result of physical pressure directly inhibiting petal primordium initiation on the sides of
the elliptic meristem or it could be that primordia initiate at
the poles of the ellipse because this is the greatest possible
distance from the center of the meristem.
Meristem shape also affects the inference of the sequence in
which primordia initiate on the meristem. Phyllotactic pattern
at later stages of bud development is characterized by particular sets of parastichy spirals. These in turn allow primordia
to be numbered from youngest to oldest. Implicit in the numbering process is that sequentially initiated primordia are arranged around the meristem at the divergence angle characteristic of the phyllotactic pattern. This appears to be the case
in flowers with more or less circular meristems. However, in
flowers with narrowly elliptic meristems, petal primordia preferentially initiate toward the poles of the meristem and only
later along the sides of the meristem. In these cases the temporal order in which positions are filled on the floral meristem
is different from that inferred from the parastichy spirals. Later
in development organ primordia have initiated over all of the
meristem surface in both terminal and lateral flowers, so that
there is no indication that the inferred order, as defined by the
divergence angle and parastichy spirals, is different from the
order in which the organ primordia were actually initiated.
Therefore, primordia in all flowers are spaced in a regular
manner, even though the process of initiation leading to that
arrangement may vary. This interpretation is supported by
simulations conducted by Douady and Couder (1996a), who
713
demonstrated that, for small plastochron ratios (such as those
displayed by flowers of D. winteri), the exact sequence of initiation of primordia does not affect the resultant phyllotactic
pattern at maturity. Thus, the disparity between the predicted
and actual sequence of organ initiation does not affect description of the phyllotactic pattern and does not prejudice the
use of divergence angle to describe the spatial arrangement of
organ primordia being initiated on the surface of the meristem.
Floral Phyllotaxis
Many different arrangements of petals, stamens, and carpels
were found, of which relatively few conformed to common
patterns such as Fibonacci phyllotaxis. Over 50% of the flowers exhibited phyllotactic patterns such as (5,6), (6,7), (7,8),
and (8,9), patterns that are rare in other angiosperms (Jean
1994). The potential for such a variety of patterns stems from
the delay between calyx and petal initiation, which appears to
release the first initiated petals from the positional cues of the
calyx. This is unlike the situation in other genera of Winteraceae, in which the position of initiated primordia affects the
position at which new primordia will initiate and where
whorled patterns are common (Doust 2000).
The first few petals initiate sequentially and form a tier
around the base of the floral meristem. These primordia are
the first members of the floral parastichies and set up the phyllotactic pattern of the flower. Initiation of the rest of the floral
organ primordia (petals, stamens, and carpels) follows without
delay and continues the phyllotactic pattern started in the basal
petal zone. The size of the petal primordia at initiation is relatively constant but that of the floral meristem is more variable.
The production of phyllotactic pattern when there is no positioning influence from previously initiated organs has been
modeled by Williams and Brittain (1984) and Douady and
Couder (1996b, 1996c). They showed that it is possible to
obtain both whorled (m,m) as well as spiral (m,m+1) patterns
for low plastochron ratios in the range observed in this study.
Whorled systems will result when the relationship between
organ primordium size and floral meristem size is such that a
whole number of organ primordia will exactly fit around the
circumference of the meristem. Spiral (m,m+1) systems will
Table 6
Significance (P) Values for Pillai’s Trace Multivariate Test for the
MANOVA of the Measured Variables Petal, Stamen,
and Carpel for the Factors Tree, Flower Position,
and Tree # Flower Position for Each
Uniflorescence Class
Uniflorescence
class
2
3
4
5
6
7
Tree
ns
ns
ns
∗
ns
∗∗
Note. ns p not significant.
∗
P ! 0.05.
∗∗
P ! 0.01.
∗∗∗
P ! 0.001.
Flower
position
Tree # flower
position
∗∗∗
ns
ns
ns
ns
ns
ns
∗∗∗
∗∗∗
∗∗∗
∗∗∗
∗∗∗
INTERNATIONAL JOURNAL OF PLANT SCIENCES
714
Table 7
Individual Tests of Significance and Approximate Variance Components (%)
for the Dependent Variables Petal, Stamen, and Carpel
Uniflorescence
class
2
3
4
5
6
7
Stamen
Petal
Tree
ns
ns
ns
ns
ns
17∗
Flower position
∗
20
ns
ns
12∗∗
12∗∗
10∗∗
Tree
Carpel
Flower position
∗∗
ns
ns
ns
ns
ns
ns
51
36∗∗
47∗∗
43∗∗
47∗∗
44∗∗
Tree
Flower position
ns
ns
ns
25∗∗
ns
ns(19∗)
10∗
ns
ns
13∗∗
ns(3∗)
6∗∗
Note. Values for the interaction term tree # flower position have been omitted, as none were significant. Significant Bonferroni-adjusted
values for each univariate ANOVA are shown with asterisks. Significance levels and approximate variance components obtained after stepdown analyses are indicated inside parentheses where they differed from univariate values.
∗
P ! 0.016 (experiment-wide critical value of P ! 0.05).
∗∗
P ! 0.003 (experiment-wide critical value of P ! 0.01).
result when the ratio of organ primordium size to floral meristem size is between values that would lead to whorls. In these
cases an imperfect whorl will be laid down first (where the
primordia are initiated at more or less the same time but not
at regular angular positions), followed by a continuous transition toward the most stable spiral mode, of the form
(m,m+1). Two thirds of the patterns observed in this study are
of the form (m,m) or (m,m+1), as, for example (5,6), (6,6),
(6,7), and (7,7). Whether a pattern will turn out to be spiraled
or whorled appears to be determined by the size of the apex
compared to the size of the primordia at initiation. Williams
and Brittain (1984) have also shown that it is possible for
higher order systems to develop from (m,m+1) systems.
A corollary of the explanation of phyllotaxis given above is
that terminal flowers, with their generally larger floral meristems, would have phyllotactic patterns with larger numbers
of parastichies than those in lateral flowers (because more petal
primordia can fit in the basal tier around the circumference of
the meristem). However, the range of phyllotactic patterns
found in lateral flowers is greater than that found in terminal
flowers, spanning parastichy sets with both fewer and more
parastichies than seen in terminal flowers. This may simply be
because there are a few lateral flowers with unusually large
floral meristems; yet there is also a correlation in lateral flowers
between the number of parastichy spirals and the shape of the
floral meristem, with the number of parastichy spirals increasing with increasingly elliptic floral meristems. This indicates
that differences in development between flowers with elliptic
versus those with more or less circular floral meristems may
account for the greater range of phyllotactic patterns seen in
lateral flowers. In flowers with circular meristems, the basal
petal primordia and the parastichies of which they are the
initial members are initiated at approximately the same time.
However, in lateral flowers with narrowly elliptic meristems,
the full set of basal primordia is not initiated until the floral
meristem has expanded enough to escape from the constraints
imposed by the tightly fitting floral bracts. Having expanded,
it is likely that more primordia can fit around the base of the
meristem than would otherwise be the case, giving rise to phyllotactic patterns with larger numbers of parastichies. Of
course, in lateral flowers with circular meristems, the basal
petal primordia initiate all around the floral meristem very
quickly once petal initiation begins, and the smaller size of the
meristem relative to terminal flowers at the same stage of development would lead to phyllotactic patterns with fewer
parastichies.
Variation in Floral Organ Arrangement within Flowers
Another component of variation in the arrangement of the
floral organs is that of variation of organ position within the
pattern shown by individual flowers. The mean divergence
angle for a flower is always close to the theoretically correct
angle for the particular pattern, but in most of the flowers
examined, divergence angles varied periodically around the
mean. This variation persisted even after the center of the meristem had been moved to its optimal position, indicating that
it is a real phenomenon of the phyllotactic pattern of the flower
rather than an artifact of the measuring process. There was a
significant positive correlation between the aspect ratio of the
floral meristem and the amount of variation in divergence angle, indicating that the periodic nature of the oscillations is a
function of the elliptic shape of the meristem.
Periodic oscillations also occurred in the values for plastochron ratio, reflecting oscillations in the distance to center of
sequentially initiated primordia. Such alternations between
longer and shorter distances to the center of the primordia can
Table 8
Number of Significant Pairwise Comparisons (%) of Floral
Organ Numbers Using the Scheffé Method
Petal
Stamen
Carpel
Trees
Lateral
flowers
Terminal and
individual lateral
flowers
Terminal and
averaged lateral
flowers
0
0
5
9
0
0
14
100
19
17
100
50
Note. Comparisons of petal, stamen, and carpel numbers were
done within each uniflorescence class, and significant comparisons
were totaled over all uniflorescence classes and given as a percentage
of the total number of comparisons.
DOUST—FLORAL VARIATION IN DRIMYS WINTERI
Table 9
Mean Differences in Floral Organ Numbers between the Terminal
and the Average of the Lateral Flowers for Each Uniflorescence
Class for the Dependent Variables Petal, Stamen, and
Carpel and Their Total for Each Flower
Uniflorescence
class
2
3
4
5
6
7
Overall mean difference
% mean increase between terminal and
lateral flowers
Petal
Stamen
Carpel
Total
1.06
0.46
0.23
0.61
0.41
0.79
0.59
6.21
6.1
5.77
6.1
6.86
6.18
6.20
0.74
0.44
!0.04
1.09
0.4
1.11
0.62
7.97
7.2
5.96
7.8
7.66
8.08
7.45
22.1a
9.23
5.8
The percentage mean increase in stamen number between terminal
and lateral flowers is significantly greater than that for petal or carpel
number. P ! 0.001.
a
be explained as the result of fitting primordia around an elliptically shaped floral meristem, with distances from the center
to the primordia on the sides of the meristem being shorter
than distances to the primordia on the poles of the meristem.
However, no significant relationship was found between variation in plastochron ratio and variation in the elliptic shape
of the meristem. This anomaly was explored by simulating the
effect of two different divergence angles on the variation in
plastochron ratio of primordia arranged around an elliptic
floral meristem. In the simulation the plastochron ratio of the
Lucas pattern (divergence angle p 99.5") varies more than that
of the Fibonacci pattern (divergence angle p 137.5") because
it is easier to obtain a short distance from the center to the
primordium (at the sides of the meristem) followed by a long
distance (at the poles of the meristem), and vice versa. The
difference in distance to center between successive primordia
for the Fibonacci pattern, and thus the variation in the plastochron ratio, is almost never as great as in the Lucas pattern.
Therefore, patterns with differing divergence angles will produce differences in the amount of variation in plastochron
ratio, even on flowers with similarly elliptic meristems. Such
confounding variation makes it impossible to meaningfully
Table 10
Comparison of Three Data Sets, Showing Differences in Petal,
Stamen, and Carpel Numbers between Terminal
and Averaged Lateral Flowers
Variable
MEL 95
difference
VUW
difference
MEL 96
difference
P values
Petal
Stamen
Carpel
0.55
6.34
0.43
!0.9
5.56
1.05
0.25
6.54
0.18
ns
ns
ns
Note. ns p not significant. P values measure significance of differences between terminal and averaged lateral flowers in petal, stamen, and carpel numbers. MEL 95, MEL 96 p collections at University of Melbourne in 1995 and 1996; VUW p collections from
Victoria University of Wellington in 1996.
715
compare variation in plastochron ratio and shape of the flower
over different phyllotactic patterns. A much larger data set is
needed to test the relationship between variability in plastochron ratio and the shape of the meristem within each phyllotactic pattern.
The mean plastochron ratio remained relatively constant
throughout the development of all flowers (line of best fit in
fig. 16B, 16D, 16F shows no significant relationship between
plastochron ratio and primordium number at P ! 0.05 level).
This indicates that primordia are produced at an even rate
throughout their initiation and that the primordia are packed
evenly and efficiently on the receptacle at initiation, presumably in order to maintain a minimum distance between themselves and other primordia. The observed variation in both
divergence angle and plastochron ratio can be interpreted to
be the result of the initiation of primordia on elliptic meristems.
The pattern that will eventuate is specific for each bud and
depends on fluctuations in the size and shape of the meristem.
Previous research on floral morphology in D. winteri has
described the arrangement of the floral organs as irregular
(Hiepko 1965), “disturbed” (Erbar and Leins 1983), or more
or less unordered (Endress 1986). Hiepko (1965) examined
12 flowers of D. winteri and interpreted the phyllotaxis as
either tending toward divergence angles of 99.5" (typical of
Lucas phyllotaxy) or of 137.5" (typical of Fibonacci phyllotaxy). However, this interpretation is probably insufficient, as
the results of this study show that many phyllotactic patterns
have divergence angles that are close to each other and to the
values of Fibonacci and Lucas patterns (table 2). In a smallscale study, Erbar and Leins (1983) foreshadowed the importance of floral meristem shape on floral organ arrangement,
although they focused on variation in organ position rather
than recognizing the diverse phyllotactic patterns that can
arise. They attributed variation in arrangement of floral organs
to asymmetry in the floral meristem but did not observe the
correlation between positional variation and meristem shape
that has been demonstrated here. Endress (1986, 1987) recognized the importance of asymmetry in producing variation
in position of floral organs and also considered that the unordered arrangement of the floral organs in D. winteri (and Winteraceae in general) was attributable to both the small size of
the primordia relative to the large floral meristem and the short
plastochron intervals. In general it has not been recognized
that the presence of diverse phyllotactic patterns is able to
explain a significant part of this variation.
Another plant system with variation in floral phyllotaxis is
the aroid spadix, and a study of 20 species by Davis and Bose
(1971) showed that 29.5% of the patterns were whorled and
71.5% spiral. Of the spiral patterns, 63.9% were Fibonacci
series, 1.3% were Lucas series, and 14.6% were of the accessory series (5,6), with 12 other accessory series making up the
balance. A significant difference between these spadices and
the flowers of D. winteri is that very few of the flowers of D.
winteri show Fibonacci phyllotaxis. Differences in phyllotactic
patterns have also been found in the heads of the sunflower
Helianthus tuberosus (Needham et al. 1993). The heads terminating the main stems of the plants showed primarily Lucas
phyllotaxy, whereas the heads terminating the lateral branches
showed Fibonacci phyllotaxy. However, in D. winteri, differences in phyllotactic patterns were not explained by the dis-
716
INTERNATIONAL JOURNAL OF PLANT SCIENCES
tinction between terminal and lateral flowers, as many patterns
were found in both types of flower. In general, the range of
phyllotactic patterns shown by flowers of D. winteri is much
larger than that shown in other plant systems. Endress (1994)
recognized that the phyllotaxis of male flowers of Ceratophyllum demersum (Ceratophyllaceae) can be whorled, irregular, or spiral, although only Fibonacci and Lucas spiral phyllotaxy were reported. It would be interesting to reexamine
these flowers to see if some of the phyllotactic patterns reported
as irregular might in fact resemble the m,m+1 patterns observed in Drimys.
Oscillations in both divergence angle and plastochron ratio
have not previously been reported in Winteraceae but have
been described in a number of other studies. Battjes et al.
(1993a, 1993b) and Battjes and Prusinkiewicz (1998) showed
that outer floret primordia in Microseris pygmaea (Asteraceae)
were initiated spirally but aligned in rings and that there were
abrupt changes in divergence angle as the spiral of sequentially
initiated primordia passed from one ring to the next. They also
showed that there were periodic oscillations in both divergence
angle and plastochron ratio (figs. 9, 15 in Battjes et al. 1993a).
They suggested that variations in divergence angles and plastochron ratio were correlated, reflecting their dependence on
the same geometrical packing rule. Ryan et al. (1991) found
periodic oscillations in divergence angle in H. tuberosus (Asteraceae) that were not removed by adjusting the center of the
meristem to minimize the variation in divergence angle. They
interpreted the remaining variation to mean that the meristem
is not patterned by a uniquely fixed center producing primordia sequentially at a fixed divergence angle but rather that
primordia pack efficiently on the meristem to create a phyllotactic pattern and its associated divergence angle.
Variation in floral morphology in D. winteri can be explained by variation in the size and shape of the meristem, the
timing of initiation of primordia, and the effect of physical
constraints on the development of pattern. The ability to explain the diversity of floral form in D. winteri by varying a
small number of developmental parameters indicates that the
similar wide diversity in floral form across the primitive an-
giosperms may be explained by the same parameters and processes. If so, the differences between the constrained small
flowers of the paleoherbs and the relatively unconstrained large
flowers of the woody magnoliids may simply reflect the ease
with which a large flower can change the relationship between
meristem size and primordium size (e.g., if primordium sizes
are equal, a 5% increase in size in a large meristem may be
enough to allow an extra primordium to be initiated, whereas
it may require a 50% increase in the size of a small meristem
to create the space for an extra primordium to be initiated).
This study indicates that variation in numbers and arrangement of floral organs is strongly affected by developmental
constraints and that direct genetic control defines the boundaries of what is possible rather than strictly determining the
actual number and arrangement of organs in individual flowers. This would have been difficult to demonstrate in highly
constrained model plant systems in which variation is minimal,
and it emphasizes the value of searching for processes controlling morphological variation in a variety of different plant
systems.
Acknowledgments
I thank Andrew Drinnan for providing the opportunity,
training, and equipment needed to accomplish this project.
Discussions with Andrew Drinnan, Bruce Sampson, Wim
Vink, Paul Green, Roger Meicenheimer, Jim Grimes, Pamela
Diggle, and Vicki Mackenzie were instrumental in forming the
views expressed above. I thank Andrew Drinnan, Elizabeth
Kellogg, Peter Stevens, Judy Jernstedt, and the Kellogg lab
discussion group for reading drafts of this article and greatly
improving it. I thank Barbara Grimes and Ingrid Foster for
German translation. Additional material of Drimys winteri
was provided by the Department of Natural Resources, Victoria, Australia, and by Bruce Sampson. Many thanks also to
reviewers Peter Endress and Dennis Barabé for their critical
and excellent suggestions.
Literature Cited
Battjes J, P Prusinkiewicz 1998 Modeling meristic characters of Asteracean flower heads. Pages 281–312 in RV Jean, D Barabé, eds.
Series in mathematical biology and medicine. Vol 4. Symmetry in
plants. World Scientific, Singapore.
Battjes J, NOE Vischer, K Bachmann 1993a Capitulum phyllotaxis
and numerical canalization in Microseris pygmaea (Asteraceae: Lactuceae). Am J Bot 80:419–428.
——— 1993b Meristem geometry and heritable variation in numbers
of florets and involucral bracts in Microseris pygmaea (Asteraceae,
Lactuceae). Acta Bot Neerl 42:255–268.
Brown C, F Mosteller 1991 Components of variance. Pages 193–251
in DC Hoaglin, F Mosteller, JW Tukey, eds. Fundamentals of exploratory analysis of variance. Wiley, New York.
Clark SE, MP Running, EM Meyerowitz 1993 CLAVATA 1, a regulator of meristem and flower development in Arabidopsis. Development 119:397–418.
Crone W, EM Lord 1993 Flower development in the organ number
mutant clavata1-1 of Arabidopsis thaliana (Brassicaceae). Am J Bot
80:1419–1426.
Davis TA, TK Bose 1971 Fibonacci systems in aroids. Fib Q 9:253–263.
Douady S, Y Couder 1996a Phyllotaxis as a dynamical self-organizing process. I. The spiral modes resulting from time-periodic iterations. J Theor Biol 178:255–274.
——— 1996b Phyllotaxis as a dynamical self-organizing process. II.
The spontaneous formation of a periodicity and the coexistence of
spiral and whorled patterns. J Theor Biol 178:275–294.
——— 1996c Phyllotaxis as a dynamical self-organizing process. III.
The simulation of the transient regimes of ontogeny. J Theor Biol
178:295–312.
Doust AN 2000 Comparative floral ontogeny in Winteraceae. Ann
Mo Bot Gard 87:366–379.
Endress PK 1978 Blütenontogenese, Blütenabgrenzung und systematische Stellung der perianthlosen Hamamelidoideae. Bot Jahrb
Syst 100:249–317.
——— 1986 Reproductive structures and phylogenetic significance of
extant primitive angiosperms. Plant Syst Evol 152:1–28.
——— 1987 Floral phyllotaxis and floral evolution. Bot Jahrb Syst
108:417–438.
DOUST—FLORAL VARIATION IN DRIMYS WINTERI
——— 1994 Evolutionary aspects of the floral structure in Ceratophyllum. Plant Syst Evol 8(suppl):175–183.
Erbar C, P Leins 1983 Zur Sequenz von Blütenorganen bei einigen
Magnoliiden. Bot Jahrb Syst 103:433–449.
Hiepko P 1965 Vergleichend-morphologische und entwicklungsgeschichtliche Untersuchungen über das Perianth bei den Polycarpicae.
Bot Jahrb Syst 84:359–508.
——— 1966 Das Blütendiagramm von Drimys winteri J.R. et G.
Forst. (Winteraceae). Willdenowia 4:221–226.
Jean RV 1994 Phyllotaxis: a systemic study in plant morphogenesis.
Cambridge University Press, Cambridge. 386 pp.
MathWorks 1992 The student edition of MATLAB. Simon & Schuster, Englewood Cliffs, N.J. 494 pp.
Minitab 1995 Version 10Xtra for Macintosh. Minitab, State College,
Pa. 430 pp.
Needham AR, JL Rouse, LA Bursill 1993 Phyllotaxis of the sunflower
Helianthus tuberosa L. Curr Top Bot Res 1:79–90.
Norusis MJ 1993 SPSS for Windows: advanced statistics, release 6.0.
SPSS, Chicago. 606 pp.
Rasmussen N, PB Green 1993 Organogenesis in flowers of the homeotic green pistillate mutant of tomato (Lycopersicon esculentum).
Am J Bot 80:805–813.
Russ JC 1995 The image processing handbook. 2d ed. CRC, Boca
Raton, Fla. 674 pp.
Rutishauser R 1998 Plastochron ratio and leaf arc as parameters of
a quantitative phyllotaxis analysis in vascular plants. Pages 171–212
717
in RV Jean, D Barabé, eds. Series in mathematical biology and medicine. Vol 4. Symmetry in plants. World Scientific, Singapore.
Ryan GW, JL Rouse, LA Bursill 1991 Quantitative analysis of sunflower seed packing. J Theor Biol 147:303–328.
Schöffel K 1932 Untersuchungen über den Blütenbau der Ranunculaceen. Planta 17:315–371.
Smith AC 1943 The American species of Drimys. J Arnold Arbor
Harv Univ 24:1–33.
Sokal RR, FJ Rohlf 1995 Biometry. WH Freeman, New York. 887
pp.
SPSS 1995 Statistical packages for the social sciences, version 6.1.3.
SPSS, Chicago.
Tabachnick BG, LS Fidell 1996 Using multivariate statistics. 3d ed.
HarperCollins, New York.
Tucker SC 1959 Ontogeny of the inflorescence and flower in Drimys
winteri var. chilensis. Univ Calif Publ Bot 30:257–336.
——— 1960 Ontogeny of the floral apex of Michelia fuscata. Am J
Bot 47:266–277.
——— 1961 Phyllotaxis and vascular organization of the carpels in
Michelia fuscata. Am J Bot 48:60–71.
Vink W 1970 The Winteraceae of the Old World. I. Pseudowintera
and Drimys—morphology and taxonomy. Blumea 18:225–354.
Williams RF, EG Brittain 1984 A geometrical model of phyllotaxis.
Aust J Bot 32:43–72.
Winer BJ, DR Brown, KM Michels 1991 Statistical principles in experimental design. 3d ed. McGraw-Hill, New York. 1057 pp.