Anim. Behav.

, 1997, 53, 1043–1052

Innate colour preferences and flexible colour learning in the

pipevine swallowtail
MARTHA R. WEISS
Center for Insect Science, University of Arizona

(Received 31 January 1996; initial acceptance 22 May 1996;

final acceptance 19 August 1996; MS. number: 7511)

Abstract. The importance of innate preferences and learned associations in choice of flower colours were
investigated for the pipevine swallowtail butterfly, Battus philenor (Papilionidae). Naive butterflies
showed innate colour preferences for yellow and, to a lesser extent, blue and purple. Apart from their
innate preferences, they were able to learn within 10 flower visits to associate floral colour with the
presence of nectar rewards in yellow or magenta Lantana camara (Verbenaceae) flowers. Continued
experience resulted in greater discrimination in favour of the rewarding colour. Most individuals
readily shifted their foraging behaviour when the colour of the rewarding flower was changed. A
capacity for rapid and flexible associative learning presumably allows butterflies to adjust their foraging
efforts in response to floral rewards that vary over space or time.
? 1997 The Association for the Study of Animal Behaviour

Ecological interactions between flowering plants
and their pollinators are the result of a long
coevolutionary history that has helped to shape
the attributes of both participants. Indirect infor-
mation about the sensory capabilities of pollin-
ators can be inferred from the existence of
‘pollination syndromes’, suites of floral traits that
can be used to characterize groups of flowers
favoured by different types of pollinators (Proctor
& Yeo 1975; Faegri & van der Pijl 1979). The
prevalence of bright red hummingbird-pollinated
flowers in diverse plant families, for example,
suggests that hummingbirds are able to perceive
that colour, and the sweet fragrance of many
moth-pollinated flowers points to that insect’s
olfactory acuity. More recently, explicit quanti-
tative analyses of the hymenopteran colour vision
system and the spectral reflectance of flowers have
demonstrated that the bee’s set of spectral recep-
tors is close to optimal for the discrimination of
floral colours (Chittka & Menzel 1992; Menzel &
Shmida 1993). In addition to sensory capacities, a
pollinator’s behavioural capabilities can also
affect the nature of its interactions with flowers.
Thus, through selection, plants are able to make
use of a pollinator’s ability to learn to extract
Correspondence: M. R. Weiss, Department of Biology,
Georgetown University, Washington, D.C. 20057-1028,
U.S.A. (email: weissm@gusun.georgetown.edu).
nectar from flowers of complex morphology
(Laverty 1980, 1994; Lewis 1986, 1993; Waser
1983), or to associate colour phase with avail-
ability of reward in colour-changing flowers (Gori
1983; Weiss 1995a). To understand the ecological
interactions between plants and pollinators, and
the evolutionary history of their relationships,
it is necessary to understand the sensory and
behavioural attributes of the pollinators.
For insects, much of our knowledge of sensory
and behavioural capabilities, including colour
vision and learning ability, comes from work on
the honey bee, Apis mellifera (von Frisch 1967;
Menzel & Erber 1978; Gould 1984; Menzel 1985,
1993). Fewer studies address colour learning
ability in the context of foraging by non-
hymenopteran insects. Work on dipterans
(Fukushi 1989; Pickens 1990) and lepidopterans
(Swihart & Swihart 1970; Swihart 1971; Lewis &
Lipani 1990; Goulson & Cory 1993; Weiss 1995b),
however, has shown that these groups of insects
are also able to learn to associate colours with
sugar rewards.
Colour vision is an important component of
intra- and inter-specific interactions in butterflies
(Crane 1955; Silberglied 1984), as well as in the
recognition and use of floral food sources. The
spectrum visible to butterflies, which extends from
the ultraviolet through the red, is the broadest

0003–3472/97/051043+10 $25.00/0/ar960357 ? 1997 The Association for the Study of Animal Behaviour
1043
1044 Animal Behaviour, 53, 5

known in the animal kingdom (Silberglied 1984).
Indeed, the japanese swallowtail, Papilio xuthus,
has five types of photoreceptor cells with different
spectral sensitivity maxima (Arikawa et al. 1987);
humans and honey bees have three.
Innate colour preferences for feeding have been
reported for a relatively broad taxonomic range
of butterflies including nymphalids, papilionids,
satyrids and pierids (Ilse 1928; Tinbergen et al.
1942; Ilse & Vaidya 1956; Swihart & Swihart
1970; Scherer & Kolb 1987a, b). A growing body
of evidence demonstrates that butterflies are also
capable of rapid learning in different contexts. In
addition to learned colour preferences in the con-
text of nectar foraging, butterflies can also learn
to associate a colour or a leaf shape with an
oviposition stimulus (Rausher 1978; Traynier
1984, 1986; Papaj & Rausher 1987). With experi-
ence, butterflies become more adept at extracting
nectar from morphologically complex flowers
(Lewis 1986) and at recognizing appropriate
larval host plants in the field (Stanton 1984).
Here I used behavioural studies to investigate
innate colour preferences and colour learning
ability in the pipevine swallowtail. I investigated
the following questions. (1) Are innate colour
preferences evident in naive B. philenor butterflies?
(2) Can B. philenor butterflies learn to associate
different flower colours with the presence of a
nectar reward? (3) Does preference for the reward-
ing colour increase with continued experience? (4)
Can a learned preference be changed in response
to new experience?
GENERAL METHODS
Study Taxa
Battus philenor
Larvae and adults of the pipevine swallowtail
were collected in the field in southern Arizona in
summer 1992 and 1993. Field-collected females
readily laid eggs on Aristolochia fimbriata plants
cultivated on the University of Arizona campus,
and larvae were raised either on fresh leaves of
A. fimbriata or on an artificial diet containing
freeze-dried A. californica leaves and shoots. In
both 1993 and 1994, a second generation was
obtained from the lab-reared butterflies. Soon
after eclosion, adults were marked on the under-
side of their hind wings with an indelible ink pen
for individual identification. Butterflies were kept
in mesh cages and hand-fed a 20% solution of
honey water twice a day. At the time of their first
experimental trial, they had had no exposure to
flowers and were considered naive. I used 43 male
and 41 female butterflies in these experiments.
Lantana camara
Flowers of L. camara (Verbenaceae) undergo
dramatic ontogenetic colour changes without
changing shape, size or turgidity (Weiss 1991;
Fig. 1). In cultivar ‘Irene’, flowers are lemon
yellow the morning they open, pink the second
morning and deep magenta the third. Changed
flowers remain fresh, and are retained on the
dense, flat-topped inflorescences for several days.
The older flowers are displaced towards the
periphery of the inflorescence as new yellow
flowers open in the centre. Inflorescences contain
a mean of 9–33% yellow flowers (Weiss 1991).
Only yellow-phase L. camara flowers produce
nectar. In the present study, unvisited yellow
flowers contained a mean& of 1.0&0.1 ìl of
nectar (N=37 flowers), at a concentration of
19.2&0.9% sucrose equivalents (N=27 flowers).
One butterfly visit is sufficient to remove all of
the nectar from a yellow flower (Weiss 1992).
Magenta flowers contain little or no nectar if they
were visited by butterflies during their yellow
phase (Weiss 1992). If yellow flowers were not
visited, however, then nectar remains in the
magenta flowers (0.91&0.1 ìl, N=45 flowers;
concentration of 17.3%&1.3 sucrose equivalents,
N=26 flowers).
EXPERIMENT 1:
INNATE COLOUR PREFERENCES
Methods
To determine relative innate colour preferences,
naive B. philenor individuals were presented with
an array of unrewarding model flowers made
of ‘Chromarama’ papers in six highly saturated
colours: yellow, blue, red, orange, green and
purple (reflectance spectra available from author).
The daisy-like flowers, approximately 6 cm in
diameter, were constructed of fringed strips of
coloured papers affixed to a central plastic pipette
tip (Crane 1955). Two flowers of each colour were
inserted into a brown cardboard background
 Weiss: Butterfly colour learning
Figure 1. Lantana camara inflorescence. Unfilled flowers
are yellow; grey flowers are magenta. Inflorescence
diameter=4 cm.
(30#45 cm); the centres of adjacent flowers were
approximately 15 cm apart, and distribution of
colours was haphazard. The array of 12 flowers
was placed on the ground inside a 2 m3 mesh field
cage. I introduced 46 naive butterflies (17 females,
29 males) individually into the cage (either
released into the air or placed on the upper wall of
the cage), and the colours of all flowers alighted
on and/or probed were recorded for each butter-
fly. The colours of the first flowers visited were
compared with (1) expected random visits, using a
1045
100
Percentage of visits
75
50
25
0
Red Orange Yellow Green Blue Purple
Colour of paper flower
Figure 2. Innate colour choices of B. philenor butterflies
on paper flowers. Naive male and female butterflies were
offered unrewarding flowers made of highly saturated
red, orange, yellow, green, blue and purple ‘Chroma-
rama’ papers. Shown are proportion of first visits (/)
and of all subsequent visits (.) to each colour.
G-test for goodness of fit, and (2) colours of
subsequent flowers visited, using a G-test of inde-
pendence. The colour choices of male and female
butterflies were also compared for all flowers
visited, using a G-test of independence.
Results
Of the six colours of paper flowers offered,
naive B. philenor butterflies showed a strong
innate preference for yellow and a secondary
preference for blue and purple. The distribution of
colours chosen in the initial visit differed signifi-
cantly from random (G5 =79.69, P<0.001). Yellow
models received approximately 67% of first visits,
and blue and purple together received approxi-
mately 30% of first visits (Fig. 2). There was no
difference between the initial and subsequent
colour choices (G5 =4.90), nor did colour choices
differ between male and female butterflies
(G5 =3.98).
EXPERIMENT 2:
LEARNED ASSOCIATIONS
Methods
I took advantage of the natural colour change
and nectar production of L. camara flowers to
produce three types of inflorescences (treatments)
that looked identical but had different patterns of
nectar distribution: ‘No Nectar’ (NN) inflor-
escences contained no nectar in either yellow or
1046 Animal Behaviour, 53, 5
magenta flowers, ‘Yellow Nectar’ (YN) inflor-
escences contained nectar in yellow but not
magenta flowers, and ‘Magenta Nectar’ (MN)
inflorescences contained nectar in magenta but
not yellow flowers. Intermediate pink-coloured
flowers were removed. Test inflorescences, like
natural L. camara inflorescences, contained
30.0&0.75% (mean&) yellow flowers (4.6&
0.17 yellow flowers; 11.3&0.61 magenta flowers;
N=86 inflorescences). The number and pro-
portion of yellow flowers in test inflorescences
were approximately equal in all treatments (NN:
4.5 yellow flowers (28.1&1.6%); YN: 5 yellow
flowers (30.9&1.1%); MN: 4.3 yellow flowers
(31.1&1.1%)). To make NN inflorescences,
butterflies were allowed to visit and remove
nectar from all flowers, including newly opened
yellow flowers. For YN inflorescences, butterflies
removed nectar from all flowers prior to the
opening of new yellow flowers. MN inflorescences
were produced by excluding butterflies from test
inflorescences for at least 3 days, so that all
flowers contained nectar. A butterfly already
experienced at visiting yellow flowers was then
held over an inflorescence and allowed to probe
yellow flowers repeatedly, but was prevented
from probing magenta flowers. Prior to the
experiments, filter paper wicks were inserted into
all non-rewarding flowers to ensure that all nectar
was gone, and a sub-sample of rewarding flowers
was examined to ensure that nectar was present.
Naive insects were placed individually on the
centre of a test inflorescence with one of the three
different nectar patterns, and the sequence of
colours of the first 10 flowers probed was
recorded.
To examine innate colour choices in the absence
of any nectar reward, 29 naive butterflies were
offered NN inflorescences. If butterflies have
an innate preference for either yellow or magenta,
then that colour should be over-visited relative
to its abundance in the test inflorescences. If
innate preference is not important, the flowers
should be visited in proportion to their abun-
dance. The colour choices of butterflies in this
treatment served as a baseline for comparison
with the choices of butterflies in the YN and MN
treatments.
I gave 29 naive butterflies YN inflorescences to
determine whether they learn to associate the
colour yellow with the presence of a nectar
reward. If so, they should over-visit yellow flowers
relative to the butterflies on NN inflorescences. If
not, their colour choices should resemble those of
the butterflies on the empty NN inflorescences.
I gave 28 naive butterflies MN inflorescences to
determine whether they learn to associate the
colour magenta with the presence of a nectar
reward. If so, they should over-visit magenta
flowers relative to the butterflies on NN inflor-
escences. If not, their colour choices should
resemble those of the butterflies on the empty NN
inflorescences.
Trials were carried out on sunny days, between
1000 and 1400 hours, inside a 2 m3 mesh field cage.
Inflorescences used in the trials were growing on
potted plants; mosquito netting was draped over
the rest of the plant to keep the butterflies away
from non-target inflorescences. In most cases, only
one inflorescence was used; in some instances two
adjacent inflorescences were used to produce the
requisite numbers of yellow and magenta flowers.
Fresh inflorescences were used for each butterfly
when flowers contained reward; in the non-
rewarding treatment, the same inflorescences were
used for two or three butterflies in some cases.
For each butterfly in all treatments, I calculated
the difference between the observed and expected
percentage of yellow flowers probed over the
course of the first 10 flower visits. The expected
value for each inflorescence was based on the
percentage of yellow flowers available in that
inflorescence. Thus a difference of zero indicates
visitation to yellow flowers in proportion to their
abundance, a positive number indicates over-
visitation, and a negative number under-visitation
to yellow. I used a two-tailed t-test to compare the
mean difference in the NN treatment with the
expected mean of zero. I used a one-way ANOVA
to compare mean differences across treatments,
and Bonferroni post hoc tests for comparison
between treatments.
To examine the development of inter-treatment
differences in choice of flower colour, I plotted,
for each of the three treatments, the mean cumu-
lative number of yellow flowers chosen at each of
the first 10 flowers visited.
Results
After only 10 flower visits, strength of prefer-
ence for yellow L. camara flowers differed
significantly between butterflies on the three
nectar treatments (F2,83 =17.13, P<0.0001; Fig. 3).
 Weiss: Butterfly colour learning 1047

Magenta nectar No nectar Yellow nectar

–40 –20 0 20 40 60 –40 –20 0 20 40 60 –40 –20 0 20 40 60

Observed – expected visits to yellow flowers (%)
Figure 3. Colour choices of naive B. philenor butterflies on three different types of L. camara inflorescences: Magenta
Nectar (nectar in magenta flowers, no nectar in yellow flowers), No Nectar (no nectar in any flowers) and Yellow
Nectar (nectar in yellow flowers, no nectar in magenta flowers). Each open diamond represents the difference between
the observed and expected percentage of visits to yellow flowers by a butterfly in its initial sequence of 10 visits;
expected values are based on the percentage of yellow flowers in the test inflorescences. Differences are ordered from
lowest to highest. A positive difference indicates preference for yellow, zero indicates no preference, and a negative
difference indicates preference for magenta. Dark diamonds on the X-axis show the mean difference for each treatment.

Butterflies on the rewardless NN inflorescences
probed a mean& of 13.93&3.39% more yellow
flowers than expected based on the abundance of
yellow flowers in the test inflorescences (t28 =4.10,
P<0.001). YN butterflies were even more dis-
criminating in favour of yellow flowers, probing
them 29.24&3.50% more than expected. MN
butterflies, on the other hand, chose yellow
flowers in proportion to their abundance (mean
difference= "0.04&3.8%). Post hoc Bonferroni
tests indicate that all treatments differed signifi-
cantly (P<0.05) from one another.
Butterflies in the three treatments differed in
both the sign and magnitude of deviation from
the expected value of zero (Fig. 3). In the MN
treatment, 17 of 28 butterflies showed negative
differences, but differences were negative for only
five and four of 29 butterflies in the NN and YN
treatments, respectively. In the YN treatment,
17 butterflies chose at least twice as many yellow
flowers as expected, but only six did so in the NN
treatment. Within each treatment, individual
butterflies varied in their choice of flower colour.
The slopes of the regressions of mean cumu-
lative visits to yellow flowers differ for the three
treatments, and indicate that treatment-specific
visitation patterns begin to develop as soon as the
butterflies begin feeding (Fig. 4). The line for the
MN treatment has a slope of 0.28, which is not
different from the slope of 0.3 that would be
expected if butterflies visited yellow flowers in
proportion to their abundance in test inflor-
escences. The steeper slopes of the NN (0.39) and
YN (0.55) treatments indicate a greater tendency
to preferentially choose yellow flowers.
EXPERIMENT 3: CHANGES IN
INDIVIDUAL PREFERENCES
Methods
To determine whether continued feeding
beyond the initial 10 visits at YN inflorescences
would result in an increased preference for yellow
1048 Animal Behaviour, 53, 5
Mean cumulative visits to yellow flowers
YN
NN
MN
5 6 7 8 9 10 1 2 3 4
Flower visited
Figure 4. Mean& cumulative number of visits to
yellow flowers at each of the first 10 flowers visited.
NN=no nectar in any flowers (N=29 butterflies);
MN=nectar in magenta flowers, no nectar in yellow
flowers (N=28 butterflies); YN=nectar in yellow
flowers, no nectar in magenta flowers (N=29 butterflies).
Dashed line is the expected number of yellow flowers
visited based on proportion of yellow flowers available.
Standard errors are symmetrical around points, but bars
are drawn in only one direction to simplify figure. Slopes
of regression lines: MN=0.28; NN=0.39; YN=0.55;
dashed line=0.30.
flowers, eight butterflies that had received YN
inflorescences when naive were allowed to
continue feeding ad libitum on YN L. camara
inflorescences for 2 additional days. Each butter-
fly’s colour choices in a sequence of 10 flower
visits were then reassessed, and each butterfly’s
initial and later colour choices were compared
using a paired t-test.
Results
All eight of the butterflies that continued to feed
on YN inflorescences for 2 additional days then
chose significantly more yellow flowers in a series
of 10 flower visits than they did as naive butterflies
(paired t-test: t7 = "8.049, P<0.0001).
EXPERIMENT 4: CHANGES IN
LEARNED ASSOCIATIONS
Methods
In this experiment, I tested whether butterflies,
having learned to associate one colour with the
presence of a reward, can switch their preference
to an alternative colour once it becomes reward-
ing. Nine butterflies trained on YN inflorescences
were subsequently offered MN inflorescences, and
nine others trained on MN were offered YN
inflorescences. Naive butterflies were placed on an
inflorescence with the initial reward treatment on
day 1; they were kept in a cage without flowers
and fed honey water on day 2, and were given an
inflorescence with the opposite treatment on day
3. Butterflies visited approximately equal numbers
of flowers in their initial training on both colours
(YN: 27.7&2.3 flowers; MN: 35.1&4.8 flowers;
two-tailed t-test: t16 =1.386, ). The colours and
sequences of all flowers probed by each butterfly
on inflorescences with each reward treatment were
recorded. To examine the transition between one
treatment and another, I plotted, for the nine
butterflies switched in each direction, the mean
percentage of yellow flowers visited while feeding
on inflorescences in the first reward scheme, and
the mean percentage of yellow flowers visited in
the first three sets of five visits on the novel reward
scheme. I used paired t-tests to compare, for each
butterfly, the percentage of yellow flowers chosen
on the initial reward scheme with the percentage
of yellow flowers chosen in the first 15 visits on the
new scheme.
Results
The percentage of visits to yellow flowers
shifted quickly and significantly following a
change in the pattern of nectar distribution in
L. camara inflorescences (Fig. 5). Butterflies
experienced on the MN reward scheme changed to
a visitation pattern characteristic of YN-trained
butterflies after only 1–5 flower visits on the new
scheme; YN butterflies took 6–10 visits to change
to a pattern characteristic of MN-trained insects
(i.e. visitation to yellow flowers in proportion
to their abundance; see Fig. 3). Paired t-tests
indicated significant differences in visitation
between baseline and novel reward schemes for
both sets of butterflies after 15 visits on the new
scheme (YN to MN, t8 =3.288, P=0.011; MN to
YN, t8 = "4.17, P=0.0031). Effects of switching
were consistent across most individuals: the
percentage of yellow flowers visited increased for
eight of nine butterflies switched from MN to YN
inflorescences, and decreased for eight of the nine
butterflies switched from YN to MN, after 15
visits on the new scheme.
 Weiss: Butterfly colour learning
100
Mean percentage of visits
to yellow flowers
50
0
BASELINE 1–5 6–10 11–15
(Initial nectar FLOWERS VISITED
treatment) (Novel nectar treatment)
Figure 5. Butterflies that were switched from one nectar
treatment to another rapidly adjust their choice of
flower colour. Nine naive butterflies were placed on an
inflorescence with one of two initial reward treatments
on day 1 (N=18 butterflies), received honey water on
day 2, and received an inflorescence with the opposite
treatment on day 3. Baseline indicates the mean percent-
age of yellow flowers visited in approximately 30 flower
visits on the initial nectar treatment. The three subse-
quent points show the mean percentage of yellow flowers
visited in the first three sets of five visits on the novel
nectar treatment. Open circles represent butterflies in-
itially trained on YN and switched to MN; dark dia-
monds represent butterflies initially trained on MN and
switched to YN. The dashed line represents the expected
percentage of visits to yellow flowers based on their
prevalence in the test inflorescences. Bars show&1 .
DISCUSSION
Stephens (1993) argued that learning should
evolve in situations in which the environment is
too unpredictable within the life of an individual
and/or over successive generations for fixed
behaviour patterns to be appropriate, but not so
unpredictable that the individual cannot behav-
iourally track its change. For nectar-feeding
insects, the relevant environmental characteristic
(availability of floral resources) is of intermediate
predictability, and many nectarivores seem to use
a combination of innate and learned behaviours
to locate and use their flowers.
Innate preferences for colours, shapes and
odours have been interpreted as providing behav-
ioural biases that aid insects in recognition and
perhaps learning of appropriate cues in the
environment (Gould 1984; Gould & Marler 1984;
Giurfa et al. 1995; Lunau & Maier 1995).
1049
Reported innate colour preferences differ between
butterfly genera and even between species. Naive
B. philenor butterflies spontaneously preferred
yellow on both paper and real flowers; a second-
ary preference for blue and purple is also apparent
on the paper flowers. These innate preferences
differ from those reported for three Papilio
species. Papilio machaon and P. demoleus pre-
ferred blue and purple coloured papers (Ilse 1928;
Ilse & Vaidya 1956); P. troilus showed a spon-
taneous feeding preference for blue papers, with
a lesser preference for orange (Swihart 1970).
Among nymphalids, innate colour preferences
also vary between species. Heliconius charitonius
spontaneously preferred orange/red followed by
blue/blue-green (Swihart & Swihart 1970).
Vanessa io and V. urticae preferred yellow and
blue equally, and V. polychlorus showed a strong
peak in yellow and a small one in blue (Ilse 1928).
Flowers pollinated by butterflies can be a wide
range of colours, including yellow, orange, pink,
red, purple, blue and mauve (Kevan & Baker
1983). Different colour preferences among other-
wise ecologically similar butterfly taxa, however,
are unlikely to reflect actual differences in preva-
lence of appropriate flower colours. An ability
to learn may make the identity of the innately
preferred colour less critical.
Beyond any innate colour preferences, an abil-
ity to rapidly associate flower colour with a nectar
reward allows individual butterflies to concentrate
their foraging efforts on a known nectar source.
As B. philenor butterflies foraged on either YN or
MN inflorescences, they learned to probe flowers
of a certain colour that provided a reward. Differ-
ent visitation patterns developed as soon as the
butterflies began feeding (Fig. 4), and after 10
visits on rewarding MN or YN inflorescences, the
butterflies’ choices of flowers differed significantly
from those of butterflies on NN inflorescences
which provided no reward. Lewis & Lipani (1989)
found that Pieris rapae could learn to associate
blue or yellow colours with a sugar reward
after only a single exposure, a rate of learning
comparable to that seen in honey bees (Menzel
1993).
Although B. philenor individuals can learn to
favour yellow flowers within 10 visits, further
foraging experience leads to an even greater pref-
erence for yellow flowers. A similar increase in
fidelity with continued rewarding experience is
seen in other insects for colour as well as other
1050 Animal Behaviour, 53, 5
learning modalities. For example, Agraulis
vanillae butterflies also become more discriminat-
ing in favour of rewarding yellow flowers (Weiss
1995b). Naive bumblebees, Bombus spp., learned
to visit morphologically complex flowers ‘cor-
rectly’ within approximately 10 visits; however,
handling times did not approach those of experi-
enced field bees until 30–60 flowers had been
visited (Laverty 1980, 1994).
Flower-colour choice by butterflies in the field
is likely to reflect an interaction between innate
and learned colour preferences. Swihart (1971)
concluded that the fundamental instinctive pat-
terns of flower-seeking behaviour in H. charitonius
are not modified by conditioning, but rather co-
exist with learned patterns. Butterflies trained to
one of their innately preferred colours (orange/red
or blue/blue-green) showed a strong peak for that
colour, but those trained to a different colour
generally visited both the innate and conditioned
colour. A similar combination of innate and
learned preferences is seen in B. philenor’s choice
of flower colour. On YN inflorescences, the pres-
ence of nectar in flowers of the innately preferred
colour yields an increased discrimination in
favour of yellow relative to the colour choices of
butterflies on NN inflorescences. The presence
of nectar in magenta flowers, on the other hand,
seems to counter-balance B. philenor’s innate pref-
erence for yellow, such that the butterflies, on
average, visit yellow flowers in proportion to their
abundance in the test inflorescences. As is clear
from Fig. 3, however, the mean difference of zero
comes about not because all butterflies on MN
inflorescences visit yellow flowers in proportion
to their abundance, but because some over-visit
yellow and others over-visit magenta. These
differences in preference do not correlate with the
sex of the butterfly.
The experimental design used in this study does
not permit a direct comparison of the rate at
which the colours yellow and magenta are
learned. Because the proportion of flowers of each
colour in the test inflorescences mimicked that
found in nature, the pattern of reward availability
was asymmetrical. Naive butterflies on YN inflo-
rescences could find nectar in approximately four
flowers, those on MN inflorescences in eight
flowers. Thus one would expect a bias towards
faster learning of magenta, based on likelihood
of encountering reward. Results of the learning
and switching trials, however, both suggest that
yellow, the innately preferred colour, is learned
more readily than magenta. Similarly, in honey
bees, the innately preferred colour is learned most
rapidly (Giurfa et al. 1995).
Learning ability allows a butterfly to respond
flexibly to spatial and temporal variation in avail-
ability of nectar flowers. Experienced B. philenor
butterflies rapidly changed their behaviour follow-
ing a shift in the pattern of nectar availability.
Changes in colour preference took place within 10
flower visits on the new reward scheme, and it is
likely that in the absence of the previously reward-
ing colour, even fewer visits would be necessary to
learn a new colour. Further, the shift was consist-
ent for most butterflies tested: 16 of 18 individ-
ually marked butterflies changed their foraging
behaviour in the direction expected within 15
flower visits. In contrast, learned preferences for
flower colour in Pieris napi were less subject to
change (Goulson & Cory 1993). After 3 days of
training to either red or blue, followed by 3 days
of conditioning to the alternative colour, 46% of
butterfly visits were still to the initially trained
colour. Goulson & Cory concluded that the per-
sistent preferences may be explained by a lower
capacity for learning in butterflies compared with
bees. Their results, however, may reflect a long
initial conditioning period, rather than a biologi-
cal difference in learning ability. Although honey
bees can readily learn to associate a new colour
stimulus with a reward after brief conditioning,
they may be unable to switch after an extended
conditioning period (Menzel 1985). Similarly,
bumblebees, that had been trained with several
hundred visits to artificial flowers of one
colour later did not sample or switch to another
colour, even though it contained superior reward
(Heinrich et al. 1977).
Results of these studies suggest that in the
absence of any new positive information,
B. philenor butterflies will continue to visit colours
that have never offered a reward or were once
but are no longer rewarding. The initial paper
flower colours chosen by naive B. philenor
foragers do not differ from subsequent colour
choices; similarly, rewards do not seem to be
necessary to maintain a conditioned preference for
yellow or magenta L. camara flowers. I have seen
both B. philenor and Agraulis vanillae butterflies,
having been conditioned to a given colour of
L. camara flower, continue to visit 50–100 or more
unrewarding flowers of that colour without
 Weiss: Butterfly colour learning
changing to the unrewarding alternative colour
(unpublished data). As occurred in the switching
experiments, however, the butterflies readily
changed to the alternative colour when they
encounter a reward.
This paper demonstrates that, apart from innate
colour preferences that presumably allow butter-
flies to recognize objects that may be flowers
against a foliage background, B. philenor
butterflies can quickly learn to associate a floral
colour with a reward. They can also readily adjust
their foraging behaviour in response to a change in
reward distribution. Such a capacity for rapid and
flexible associative learning would allow butterflies
to adjust their foraging efforts in response to floral
rewards that vary over space or time.
ACKNOWLEDGMENTS
I thank the Center for Insect Science for financial
and logistical support. E. A. Bernays and D. R.
Papaj provided equipment, advice and assistance
with butterfly raising. E. A. Bernays, D. Carr, R.
Chapman, L. Chittka, J. Glendinning, E. Nagy,
J. P. Rosenthal and two anonymous referees
offered constructive comments on the manuscript.
Mark Tater generously shared his recipe for
Battus diet and provided Aristolochia californica
powder. The American Association of University
Women provided support, and M. Dudash and
C. Fenster at the Department of Plant Sciences at
the University of Maryland space, while I was
writing the manuscript. To all I am grateful.
REFERENCES
Arikawa, K., Inokuma, K. & Eguchi, E. 1987. Penta-
chromatic visual system in a butterfly. Naturwissen-
schaften, 74, 297–298.
Chittka, L. & Menzel, R. 1992. The evolutionary adap-
tation of flower colours and the insect pollinators’
colour vision. J. comp. Physiol. A, 171, 171–181.
Crane, J. 1955. Imaginal behavior of a Trinidad butter-
fly, Heliconius erato hydara Hewitson, with special
reference to the social use of colour. Zoologica
(N.Y.), 40, 167–196.
Faegri, K. & van der Pijl, L. 1979. The Principles of
Pollination Ecology. Oxford: Pergamon Press.
von Frisch, K. 1967. The Dance Language and Orien-
tation of Bees. Cambridge, Massachusetts: Harvard
University Press.
Fukushi, T. 1989. Learning and discrimination of
coloured papers in the walking blowfly, Lucilia
cuprina. J. comp. Physiol. A, 166, 57–64.
1051
Giurfa, M., Nunez, J., Chittka, L. & Menzel, R. 1995.
Colour preferences of flower-naive honey bees.
J. comp. Physiol. A, 177, 247–259.
Gori, D. F. 1983. Post-pollination phenomena and
adaptive floral changes. In: Handbook of Experimental
Pollination Biology (Ed. by C. E. Jones & R. J. Little),
pp. 31–49. New York: Scientific and Academic
Editions.
Gould, J. L. 1984. Natural history of honey bee learn-
ing. In: The Biology of Learning (Ed. by P. Marler &
H. S. Terrace), pp. 149–180. Dahlem Konferenzen
1984. Berlin: Springer-Verlag.
Gould, J. L. & Marler, P. 1984. Ethology and the
natural history of learning. In: The Biology of
Learning (Ed. by P. Marler and H. S. Terrace),
pp. 47–74. Dahlem Konferenzen 1984. Berlin:
Springer-Verlag.
Goulson, D. & Cory, J. S. 1993. Flower constancy and
learning in foraging preferences of the green-veined
white butterfly Pieris napi. Ecol. Entomol., 18,
315–320.
Heinrich, B., Mudge, P. R. & Deringis, P. A. 1977.
Laboratory analysis of flower constancy in foraging
bumblebees: Bombus ternarius and B. terricola. Behav.
Ecol. Sociobiol., 2, 247–265.
Ilse, D. 1928. U } ber den Farbensinn der Tagfalter.
Z. vergl. Physiol., 8, 658–692.
Ilse, D. & Vaidya, V. G. 1956. Spontaneous feeding
response to colours in Papilio demoleus L. Indian
Acad. Sci. Proc. B, 43, 23–31.
Kevan, P. G. & Baker, H. G. 1983. Insects as flower
visitors and pollinators. A. Rev. Entomol., 28, 407–
453.
Laverty, T. M. 1980. The flower-visiting behaviour of
bumble bees: floral complexity and learning. Can J.
Zool., 58, 1324–1335.
Laverty, T. M. 1994. Bumble bee learning and flower
morphology. Anim. Behav., 47, 531–545.
Lewis, A. 1986. Memory constraints and flower choice
in Pieris rapae. Science, 232, 863–865.
Lewis, A. C. 1993. Learning and the evolution of
resources: pollinators and flower morphology. In:
Insect Learning: Ecological and Evolutionary Perspec-
tives (Ed. by D. R. Papaj & A. C. Lewis), pp. 219–242.
New York: Chapman & Hall.
Lewis, A. C. & Lipani, G. 1990. Learning and flower use
in butterflies: hypotheses from honey bees. In: Insect–
Plant Interactions, vol. II (Ed. by E. A. Bernays),
pp. 95–110. Boca Raton, Florida: CRC Press.
Lunau, K. & Maier, E. J. 1995. Innate colour prefer-
ences of flower visitors. J. comp. Physiol. A., 177,
1–19.
Menzel, R. 1985. Learning in honey bees in an
ecological and behavioral context. In: Experimental
Behavioral Ecology and Sociobiology (Ed. by B.
Hölldobler & M. Lindauer), pp. 55–74. New York:
Fischer Verlag.
Menzel, R. 1993. Associative learning in honey bees.
Apidologie, 24, 157–168.
Menzel, R. & Erber, J. 1978. Learning and memory in
bees. Scient. Am., 239, 102–111.
Menzel, R. & Shmida, A. 1993. The ecology of flower
colours and the natural colour vision of insect
1052 Animal Behaviour, 53, 5
pollinators: the Israeli flora as study case. Biol. Rev.,
68, 81–120.
Papaj, D. R. & Rausher, M. D. 1987. Components
of conspecific host-plant discrimination by Battus
philenor (Papilionidae). Ecology, 68, 245–253.
Pickens, L. G. 1990. Colorimetric versus behavioral
studies of face fly (Diptera: Muscidae) vision. Environ.
Entomol., 19, 1242–1252.
Proctor, M. & Yeo, P. 1975. The Pollination of Flowers.
London: Collins.
Rausher, M. D. 1978. Search image for leaf shape in a
butterfly. Science, 200, 1071–1073.
Scherer, C. & Kolb, G. 1987a. Behavioral experiments
on the visual processing of color stimuli in Pieris
brassicae L. (Lepidoptera). J. comp. Physiol. A, 160,
645–656.
Scherer, C. & Kolb, G. 1987b. The influence of color
stimuli on visually controlled behavior in Aglais
urticae L. and Pararge aegeria L. (Lepidoptera).
J. comp. Physiol. A, 161, 891–898.
Silberglied, R. E. 1984. Visual communication and
sexual selection among butterflies. In: The Biology of
Butterflies (Symposium of the Royal Entomological
Society of London, number 11), (Ed. by R. I. Vane-
Wright & P. R. Ackery), pp. 207–223, London:
Academic Press.
Stanton, M. L. 1984. Short-term learning and the
searching accuracy of egg- laying butterflies. Anim.
Behav., 32, 33–40.
Stephens, D. W. 1993. Learning and behavioral ecology:
incomplete information and environmental predict-
ability. In: Insect Learning: Ecological and Evolution-
ary Perspectives (Ed. by D. R. Papaj & A. C. Lewis),
pp. 195–218. New York: Chapman & Hall.
Swihart, C. A. 1971. Colour discrimination by the
butterfly, Heliconius charitonius Linn. Anim. Behav.,
19, 156–164.
Swihart, C. A. & Swihart, S. L. 1970. Colour selection
and learned feeding preferences in the butterfly,
Heliconius charitonius Linn. Anim. Behav., 18, 60–64.
Swihart, S. L. 1970. The neural basis of colour vision in
the butterfly, Papilio troilus. J. Insect Physiol., 16,
1623–1636.
Tinbergen, N., Meeuse, B. J. D., Boerema, L. K. &
Varossieau, W. W. 1942. Die Balz des Samfalters
Eumenis (Satyrus) semele L. Z. Tierpsychol., 5, 1.
Traynier, R. M. M. 1984. Associative learning in the
ovipositional behaviour of the cabbage butterfly,
Pieris rapae. Physiol. Entomol., 9, 465–472.
Traynier, R. M. M. 1986. Visual learning in assays of
sinigrin solution as an oviposition releaser for the
cabbage butterfly, Pieris rapae. Entomol. exp. appl.,
40, 25–33.
Waser, N. M. 1983. The adaptive nature of floral traits:
ideas and evidence. In: Pollination Biology (Ed. by
L. Real), pp. 241–285. Orlando: Academic Press.
Weiss, M. R. 1991. Floral colour changes as cues for
pollinators. Nature, Lond., 354, 227–229.
Weiss, M. R. 1992. Ecological and evolutionary signifi-
cance of floral color change. Ph.D. dissertation,
University of California, Berkeley, California.
Weiss, M. R. 1995a. Floral color change: a widespread
functional convergence. Am. J. Bot., 82, 167–185.
Weiss, M. R. 1995b. Associative colour learning in a
nymphalid butterfly. Ecol. Entomol., 20, 298–301.