The Condor98:7&Z-790 Q TheCooper Ornithological Society 1996

FRUIT COLOR CHOICES OF CAPTIVE ( LATERALIS’)’

HELEN L. PLJCKEY, ALAN LILLY AND DENNIS J. ODOWD~’ Department of Ecology and Evolutionary Biology, Monash University,Clayton, 3168,

Abstract. Fleshy fruits occurin many colorsin nature,but red and blackpredominate. One popularhypothesis to explain the adaptivesignificance of fruit colorationis that it attractsfrugivorous that disperseseeds. We presentedSilvereyes (Zosterops lateralis), importantfruit consumersin southernAustralia, with choicesin the aviary betweenboth artificialfruits (made from gelatin)and actualfruits (Rhagodia parabolica) of threedifferent colors(red, yellow and white).Silvereyes exhibited a strongoverall preference for redamong both artificialand real fruits.Although individual birdsvaried in their colorpreferences for artificial fruits, all preferredthe red fruits of R. parabolica. The consumptionrate of real fruits was much greaterthan that of artificial fruits, which was probablyattributable to differencesin the characteristicsof the two fruit types.Exposure of Silvereyesto a main- tenancediet of a particularcolor for 12 daysalso failed to alter their collectivepreference for red fruits. Further, responsesto artificial infmctescencesoffering color choiceseither betweenor within infmctescencesshowed that Silvereyespreferred red fruits regardlessof the spatialformat of presentation.The overall fruit color preferencesof Silvereyeswere basedon selectionof huerather than brightness,but within the samehue, some individuals exhibitedpreferences for particularbrightness levels. These results are consistentwith the hypothesisthat fruit color is relatedto avian frugivoryand suggestthat birds can act as strongselective agents on fruit color. Key words: brightness;color; frugivory; fruit-color polymorphism;fnrits; hue;preference; ;Zosterops lateralis.

INTRODUCTION in their initial color preferences,and in the tran- Red is one of the most common colors of - sitivity and temporal stability of these prefer- dispersed fruits (Ridley 1930). Surveys of re- ences. Often, however, inferences about the im- gional floras, despitetheir taxonomic differences, portance of color in determining fruit preferences support this generalization (Turcek 1963, Knight in aviary studieshave been confounded by cross- and Siegfried 1983, Gautier-Hion et al. 1985, speciescomparisons of fruits that introduce oth- Wheelwright and Janson 1985, Lee et al. 1988, er, uncontrolled variables suchas their size,shape, Willson et al. 1989). Willson and Whelan (1990) taste and nutritional value (Turcek 1963, Mc- proposeda set of hypothesesto explain the evo- Pherson 1987, Moermond et al. 1987). Further, lution of fruit color. Of these, perhaps the most in studies involving artificial “fruits”, it is un- intuitively appealing hypothesis is that certain clear whether their use reflects the birds’ pref- colors are more attractive to frugivorous birds erencesfor real fruits (Willson et al. 1990). Field- and promote seeddispersal. However, only a few based studies on the influence of color on fruit experiments have examined whether birds dis- selection are complicated by other factors such criminate among fruits on the basis of color (Mc- as crop size, accessibility, relative abundance, Pherson 1987, 1988, Willson et al. 1990, Willson and differences in plant morphology (Wheel- and Comet 1993, Willson 1994). Collectively, wright and Janson 1985, McPherson 1987, Will- these studies showedthat individual birds differ son and O’Dowd 1989, Whelan and Willson 1994). Silvereyes (Zosterops lateralis, Zosteropidae) ’ Received19 March 1996.Accepted 25 June 1996. are major consumers of fleshy fruits in southern z Correspondingauthor. Address for all correspon- Australia, consuming fruits of just under half of dence:Department of Ecologyand EvolutionaryBi- the 100 or so plant specieswhose fruits are re- ology, Monash University, Clayton, Victoria 3 168, ported to be taken by birds in temperate Aus- Australia, e-mail: [email protected] 3Current address: Institute of Pacific Islands For- tralia (Forde 1986, French 1990). We used aviary estry,Forestry Research Laboratory, 1643Kilauea Ave., experiments to determine whether Silvereyes ex- Hilo, HI 96720. hibit a distinct color preferencein their selection

[7801 FRUIT COLOR CHOICE BY SILVEREYES 781 of fruits. First, we examined the strength and we determined preferencesby the number of fruits consistency of their fruit-color preferencesover of each color consumed and the order in which a short time scale,using both artificial fruits and they were consumed. Once presented with the the polymorphic fruits of the shrub, Rhagodia test fruits, eachbird was observedat 2 m distance parabolica R.Br. (Chenopodiaceae).Second, be- from behind a blind. Using an event recorder, cause birds may learn quickly to change their we monitored the number of fruits of each color preferencesfor fruits, depending on their degree taken and the sequencein which they were re- of exposure to them and on their availability moved over a specified time. (Morden-Moore and Willson 1982) we exposed The use of gelatin-based, artificial fruits as well Silvereyes to a diet of a particular color to see if as the natural polymorphic fruits of Oldman their fruit color preferenceschanged as a result. Saltbush, Rhagodia parabolica, allowed us to Third, the spatial scale at which fruit color se- compare color preferences involving different lection might be made is poorly known. We ex- pigment systems and attempt to control for all amined the effect of scale of fruit presentation factors other than color itself (Appendix 1). Ge- (within and between artificial infructescences)on netically polymorphic fleshy fruits occur where the color preferences of Silvereyes. Lastly, we different individuals of the same speciesproduce determined whether the fruit color preferences fruits of different colors. Individual plants of of Silvereyes were based on differencesin hue or Rhagodiaparabolica produce red, yellow or white brightness, the two aspectsthat make up what fruits and these color morphs do not differ sig- we commonly refer to as color (Goldstein 1989). nificantly in size, mass, pulp-seed ratio, water content or major nutrients (Willson and O’Dowd METHODS 1989). Artificial fruits were made from a sugar and EXPERIMENTAL PROCEDURES gelatin recipe modified from Levey and Grajal Twenty-seven adult Silvereyes were caught un- (199 1) and were dyed either red, yellow (using der permit in mist nets between 29 February McKormickO food dyes) or white (using titani- 1993 and 7 January 1994 at Bacchus Marsh, um white) to approximate the same color stan- approximately 53 km west of Melbourne, Vic- dards as the fruits of R. parabolica (approximate toria, Australia (37”37’S, 144”25’E). The birds Methuen colors: lOB8, 3A7 and lA1, respec- were maintained in a holding aviary (approxi- tively, Komerup and Wanscher 196 1). Fruits of mately 3 x 2 x 2 m). They were given at least R. parabolicawere collectedin January 1993 from two weeks to adjust to captive conditions and remnant eucalypt mallee at Djerriwarrh Creek were fed on a maintenance diet of FarexO baby (Myers et al. 1986) approximately 11 km south- food, apples, pears, and water. For all experi- east of Bacchus Marsh. Fruits were frozen at ments, birds were placed in individual cages(37 - 15°C to prevent deterioration and to enable cm wide x 50 cm deep x 36 cm high) and given experiments to be carried out when the fruits an additional 5 days on the maintenance diet to were not available in the field. adjust. These cages were placed together in a room such that the birds could not make visual EXPERIMENT 1: COLOR SELECTION OF contact, but could hear one another. Each cage ARTIFICIAL AND REAL FRUITS had a wire mesh front that allowed accessand Fruit-color preferences of Silvereyes were first observations to be made. Percheswere placed at examined using artificial fruits and then fruits of both ends of each cageand food was always pre- R. parabolica, with an interval of approximately sented at the front of the cage. On any test day, one week between test series.Fourteen birds were the birds were tested serially after being deprived testedonce daily for seven consecutivedays, and of food for one hour; after the trial, each bird in each trial all three fruit colors were presented was placed on the maintenance diet again. Water simultaneously. Each trial lasted for 25 minutes. was available ad libitum. The trials in any one Three (10 cm diameter) glass petri-dishes were experiment were carried out on consecutivedays placed in a row on the floor at the front of the between 06:00-ll:OO. The time at which each cage. This forced the birds to leave the perches bird was tested was varied systematically to test to feed. Each dish contained 20 fruits of a single for any possible effect of time of day on fruit- color. They were presentedon a rectangular(34.5 color preferences.Like Moermond et al. (1986), x 10 cm) background of green cardboard (Me- 782 HELEN L. PUCREY ET AL.

thuen 26E8) to approximate the contrast be- the first type of presentation. The secondtype of tween the fruits and their background color in presentation was then administered in the same nature. The positions in which the different col- way, after a one-day interval. ored fruits were presented in each trial were se- EXPERIMENT 4: HUE AND lected randomly from the six different permu- BRIGHTNESS tations possible to control for any effect of po- sition. Using the same birds as in Experiment 3, we examined whether the apparent color prefer- EXPERIMENT 2: PREVIOUS COLOR ences of Silvereyes are based on hue or bright- EXPERIENCE ness. Hue is determined by the spectral com- Fruit removal before and after exposureto a col- position of the light reflected from the stimulus. ored maintenance diet was compared to deter- Brightness, or luminance, is determined by the mine if previous experience affected fruit-color total amount of energy over all wavelengths re- preferences.Fifteen Silvereyes were tested, 12 of flected from the stimulus. If a bird, presented which had been used previously in Experiment with its preferred fruit color in conjunction with 1. Birds were fed on the standard maintenance a seriesof gray fruits, ranging from white to black, diet in the aviary and test cage prior to the ex- still takes its favored color, it respondsto hue as periment. They were then given fruit-color pref- such (Muntz 1974). A series of artificial fruits erence tests with R. parabolica for 15 minutes were made from the gelatin recipe by adding dif- each on four consecutive days. The birds were ferent amounts of black and white food dye to then divided randomly into three groups of four approximate a gray scaleof O%, 25%, 50%, 75%, (the three birds that had not participated in Ex- and 100% of black. Measurement with a pho- periment 1 were allocated so that one was in each tographic spot-meter (CalculightO) gave relative of the three groups). Over the next 12 days, one log,, values of luminance for the fruits of 1.00, of the groups was fed on the original whitish 0.63, 0.32, 0.20, and 0.06, respectively. Thus, maintenance diet, while the second and third these fruits differed in brightness, but not hue. groups were fed exactly the same maintenance Red artificial fruits of two brightnesslevels (0.06 diet, except that it had been dyed either red or and 0.10 log,, values as measured with the spot- yellow with McKormickO food dyes. After this meter) were also used to control for the possi- period of exposure, the birds were re-tested for bility that the birds may selectfruits on absolute another four days. brightnessvalues. Ten artificial fruits were placed in each of seven (50 mm diameter) plastic petri- EXPERIMENT 3: SPATIAL SCALE dishes, such that each dish contained only one Twelve experimentally naive Silvereyes were fruit type. Petri-dishes were assignedrandomly testedto determine if scaleofpresentation affects to positions. Each bird was tested daily for 15 fruit-color choice. Artificial infructescenceswere minutes over four consecutive days. constructed with florist’s materials: green, plas- tic-coated wire “stems” and small green cloth STATISTICAL ANALYSES “leaves”. On each infiuctescence, 15 fruits of R. The simplest and most conservative way to as- parabolicawere mounted on the ends of the stems sessthe preferencesof Silvereyes when all fruits against a background of seven leaves. During were equally available was to record only the trials, the infructescences were hung from the color of the first fruit taken in each trial (Table wire front of the cage, approximately 5 cm from 1). However, fuller insight can be gained by con- the floor. The artificial infructescenceswere pre- sidering the number of fruits taken over some or sented to Silvereyes in two ways. First, three in- all of the trial. The counts were analyzed with a fructescences,each with 15 fruits of a single color log-linear model assuming that the errors were (red, yellow or white), were presented simulta- Poissondistributed. The log-likelihood ratio sta- neously in trials. In this way, birds effectively tistic (LRS) was used to assessthe fit of the model selectedfruit colors by choosing among infruc- and to test any hypothesesusing GLIM (General tescences.Second, five fruits of each color were Linear Interactive Modelling, see Aitkin et al. arranged haphazardly on each of the three in- 1987, Crawley 1993). The LRS is chi-square dis- fructescencessuch that birds selectedfruit colors tributed (Aitkin et al. 1987). within infructescences.Each bird was tested dai- Two inherent biases in the data were caused ly for 15 minutes on four consecutive days on by the tendency for fruit of a particular color to FRUIT COLOR CHOICE BY SILVEREYES 783

TABLE 1. Fruit-color preferencesbased on the first fruit consumed by Silvereyes in each trial. Proportions are based on the number of trials in which a particular color was consumed first. Values in parenthesesare the total number of trials. Experiments are listed in chronologicalorder. Birds that did not respond to fruits are excluded.

Proportionof first fruits consumed Experiment Red Yellow White No. birds; triakhiid 1a. Artificial fruits 0.62 0.06 0.32 11; 7 (77) lb. R. parabolica 0.85 0.09 0.06 13; 7 (91) 2. Previous exposure Pre 0.77 0.08 0.15 15; 4 (60) Post 0.70 0.13 0.17 15; 4 (60) 3. Spatial scale Within infiuctescence 0.73 0.04 0.23 12; 4 (48) Between infructescence 0.69 0.12 0.19 12; 4 (48) be taken in runs and by fruit depletion. A bird already perching on a dish is more likely to take fruit from that dish than to move to another one. Thus, the trials tend to result in clustering (“ov- (a) Artificial fruits erdispersion”) of data. This can be overcome with GLIM by introducing a scalingfactor, found by dividing the residual deviance by the residual degreesof freedom (Aitkin et al. 1987). The anal- ysis is first run as a Poisson model and then the scaling factor is derived and the model re-run. Because we did not replace fruits removed by birds, the probability of taking the favored color necessarilydecreased during the trial. This would decreasethe likelihood of detecting a real pref- Red Yellow White erence. Except in Experiment 4, a cut-off point when 50% of fruits had been taken was therefore (b) Rhagodiaparabolica used, rather than a specified elapsed time, be- cause birds depleted fruits at different rates. A z IO- lower cut-off point would tend to exclude infor- 3 mation about the frequencies at which the less z15- T 1 e preferred fruits were taken. a * Preferences are expressed as the number of Ew IO- fruits of different colors consumed in the feeding 2 * trials or as a rank based on the relative propor- 5 5 0 tions in which the different fruits were consumed. c ; o_ GLIM analysesindicate the significanceof effects and interactions and, by inspection of the data, Red Yellow White we can rank the colors in a preferenceorder. Box FIGURE 1, Boxplots of the number of fruits of each plots (Wilkinson 1990) rather than means (f SE) color taken by Silvereyesper trial for (a) artificial fruits are used to graphically portray the results, be- (excludingthree birds that did not respond to the test causethe distributions of fruit removal were non- fruits) and (b) fruits of Rhagodiaparabolica (excluding one non-respondingbird). The boxplot showsthe me- normal and had heterogeneousvariances. dian (cross bar in open box) and 25% quartiles (rep- resented by each end of the box). The length of the RESULTS entire box or “spread” representsthe interquartilerange. EXPERIMENT 1: SELECTION OF The line extending from each end of the box encom- ARTIFICIAL AND REAL FRUITS passesall values within a range beyond the 25% quar- tile that is 1.5 times the spread. The stars represent Given a choice between red, yellow or white ar- outliers and open circles represent far outside values tificial fruits, Silvereyes showed an overall pref- (see Wilkinson 1990). 784 HELEN L. PUCKEY ET AL..

TABLE 2. GLIM analysis of the color preferencesof Silvereyes for both the artificial fruits (excluding three non-respondingbirds) and the fruits of Rhugodiaparabolica (excludingone non-respondingbird). Analyseswere based on the first 50% of fruits (30 fruits) consumedby individual Silvereyesin each trial. ns = P > 0.05.

Artificialfruits R. parabolica SOUICe X' df P X2 df P Bird 17.0 10 ns 13.5 12 ns Color 275.0 2 -=z0.001 520.9 2 < 0.001 Day 2.2 1 ns 4.5 1 < 0.05 Bird x day 13.8 12 ns Color x bird 37:.: :oo < ol”oo1 240.4 24 < 0.001 Color x day 18:4 2 < 0.01 3.5 2 ns Bird x color x day 51.3 20 < 0.001 58.7 24 < 0.001

erence ranking of R> W >Y (Fig. la). This pref- = 2.8, df = 6, P > 0.05) nor dish position (for erence for red was highly significant (color effect, artificial fruits, x2 = 1.4, df = 5, P > 0.05; for Table 2) and was consistent with the strong pref- R. parabolica, x2 = 1.5, df = 5, P > 0.05) had erence for red observed when only the first ar- any significant effect on fruit color choice. tificial fruit consumed in each trial was consid- ered (Experiment la, Table 1). Still, a few birds EXPERIMENT 2: PREVIOUS COLOR deviated in either their color preferenceor in the EXPERIENCE consistency of their preferences over the 7-day Following exposure to a maintenance diet of a period (color x bird, color x bird x day inter- single color (either red, yellow or white) for 12 actions; Table 2). Of the 11 of 14 birds that con- consecutive days, Silvereyes showed no signifi- sumed artificial fruits, eight preferred red and cant overall change in their color preferences three preferred white. (color x pre/post interaction, Table 3) and still Silvereyes showed an overall preferencerank- chose red fruits of R. parabolica (Fig. 2). This ing of R>Y>W when given the same color was consistent with analyses based on the first choicesof natural R. parabolica fruits (Fig. lb). fruit consumed (Experiment 2, Table 1). Groups This preference for red was highly significant exposedto white and yellow maintenance diets (color effect, Table 2) and consistent with the both showed a fruit color preference ranking of strongpreference for red apparent when only the R>W>Y before and after exposure to the diet first fruit consumed in each trial was considered (Fig. 2a,b). For Silvereyesexposed to a red main- (Experiment lb, Table 1). Although the 13 re- tenance diet, preferencesranked R>Y > W pre- sponding birds all preferred red fruits, individual and post-exposure(Fig. 2~). These differencesin Silvereyesdiffered in the relative strengthof their rankings of white and yellow fruits were reflected preferencesfor yellow and white (color x bird in the significant interaction between color and interaction, Table 2). exposure group (Table 3). However, no differ- For both fruit types, neither time of day at ence was found for any group in the number of which the birds were tested (for artificial fruits, fruits of each color eaten before and after they x2 = 3.7, df = 6, P > 0.05; for R. parabolica, x2 were exposed to a particular colored mainte-

TABLE 3. GLIM analysisofthe effectsofexposure to a single-coloredmaintenance diet on the color preferences of Silvereyes for fruits of Rhagodia parabolica. ns = P > 0.05. Analyses were based on the first 50% of fruits (30 fruits) consumed by individual Silvereyesin each trial.

!3ource X2 df P Color 138.52 2 < 0.001 Pre/post exposure 0.73 1 ns Exposuregroup 0.17 2 ns Color x pre/post 2.47 2 Color x exposuregroup 33.98 4 < ol”oo1 Pre/post x exposuregroup 0.29 2 ns Color x pre/post x exposuregroup 6.30 4 ns FRUIT COLOR CHOICE BY SILVEREYES 785

(a) Exposure to White

Red Yellow White Red Yellow White Pre-exposure Post-exposure

(b) Exposure to Yellow

Red Yellow White Red Yellow White Pre-exposure Post-exposure

I Exposure to Red

Red Yellow White Red Yellow White Pre-exposure Post-exposure

FIGURE 2. Boxplotsof the numbers of red, yellow and white fruits of Rhagodiaparubolicutaken by Silvereyes pre- and post-exposureto a (a) white maintenance diet, (b) yellow maintenance diet, and (c) red maintenance diet. nance diet (color x pre/post x exposure group This order was also apparent in the color of the interaction, Table 3). first fruit consumed in each trial (Experiment 3, Table 1). Analyses showed a significant overall EXPERIMENT 3: SPATIAL SCALE color preference (color effect, Table 4) but no Scale of presentation of R. parabolica fruit did significant difference in the color preferencesin not affect the overall color preferencesof Silver- the two choice situations (color x betweemwith- eyes. In both choice situations (between and in interaction, Table 4). However, some indi- within infructescences),red was preferred, fol- vidual birds showed significantly different color lowed by white and then yellow fruits (Fig. 3). preferencesin the two presentation types (bird 786 HELEN L. PUCKEY ET AL.

(a) Within inftuctescences TABLE 4. GLIM analysis of the preferencesof Sil- vereyes for fruits of Rhagodia parabolica at two dif- ferent spatial scales (within and between infructes- zLI cences).One bird was not included in the analysisbe- _= causeit did not take fruits in either situation. ns = P -z > 0.05. Analyses were based on the first 50% of fruits 3 10 (22 fruits) consumed by individual Silvereyesin each ; trial.

$ Source xi df P 5 15: +, f Bird 8.12 11 ns v1 .s Between/within 0.25 1 ns 1.17 3 & P Day 0 Color 106.57 2 < &Ol Bird x between/within 2.16 11 ns Red Yellow White Bird x day 5.57 33 ns Color x between/within 0.95 2 (b) Between infiuctescences Day x color 13.01 6

in brightness(Fig. 4). This result was reflectedin the proportions of fruits removed first. In 92 percent of trials, Silvereyes took red fruits first; in the others, white was chosen first. We only analyzed variation in the number of red fruits of each brightness level consumed because Silver- eyestook few other fruits. Brightnesshad no sig- nificant overall effect on removal of these red fruits (x2 = 0.1, df = 1, P > 0.05, brightness Red Yellow White level); this was consistent with the proportions FIGURE 3. Boxplots of the numbers of red, yellow, of fruits consumed first by Silvereyes, namely and white fruits of Rhagodia parabolica taken by Sil- 50% and 42% for higher and lower brightness vereyes per trial (excluding one non-respondingbird) levels of red fruits, respectively. However, in- when the choiceof colorswas presentedeither (a) with- dividuals varied in their preferencesfor different in or (b) between artificial infructescences. shades of red (x2 = 31.7, df = 10, P -C 0.001, bird x brightnessinteraction). There was no sig- x color x b/w interaction, Table 4). Of the 11 nificant daily variation in fruit removal (x2 = birds that took fruits, seven showed a consistent 1.8, df = 3, P > 0.05, effect of day) or in the preference for red fruits, irrespective of scale of total number of fruits eaten by individual birds presentation. Three birds preferred red fruits at in each trial (x2 = 4.6, P > 0.05, bird x day the within-infructescence scale but white fruits interaction). between infructescences; one individual took DISCUSSION more red fruits at the within-infructescence scale, CONSISTENCY IN FRUIT COLOR SELECTION but more yellow from the other format. Red color was imnortant in the fruit choices of Silvereyes in our aviary experiments. First, red EXPERIMENT 4: HUE AND BRIGHTNESS was preferred overall for both artificial and R. Silvereyes selectedfruits on the basis of hue rath- parabolica fruits (Experiment I), even though the er than brightness. Given a choice of artificial actual pigments of the fruit types differed. Sec- fruits of five brightness levels of gray and two ond, the attraction of Silvereyes to red fruits was brightness levels of red, the birds exhibited a not altered by exposure to a particular colored strong overall preferencefor red. Of the 11 birds maintenance diet (Experiment 2). Third, Silver- that consumed fruits, 10 ate red fruits of both eyes preferred red fruits regardlessof the spatial brightnesslevels almost exclusively, and one ate scale (within and between infructescences) or white fruits and the gray fruits closest to them mode of the presentation (petri dish or infruc- FRUIT COLOR CHOICE BY SILVEREYES 787

3 10 -) 0 h 0 ; 8- 2 0 ; 6- 8 4- 0

% s v) 2- 0 .c Lz 0 0 0 I I I I I I Red1 Red2 White Greyl Grey2 Grey3 Black FIGURE 4. Number of artificial fruits taken by Silvereyesper trial when given a choiceof two brightness levelsof red (1 and 2) and five shadesof gray (includingblack and white). Data are for the entire 15 minute trials. tescence)employed (Experiment 3, Experiment The suggestion that fiugivorous birds can 1 vs. Experiment 3). Together, theseexperiments quickly change their fruit-color preferencesde- show a temporal and spatial consistencyin color pending on exposure (Morden-Moore and Will- selection not reported in aviary studies of other son 1982) was not supportedin our study. There- frugivorous birds. Captive Northwestern Crows fore, Silvereyes’ fruit-color preferencesprobably (Corvuscaurinus) and American Robins (Turdus were not affected by experience with particular migratorius) exhibited considerable individual fruit colors prior to capture. The use of white as variation in their color preferences when pre- an exposure color in one group of birds con- sentedwith red, blue, yellow, and green artificial trolled for any effect of previous experiencewith fruits (Willson and Comet 1993, Willson 1994). the white maintenance diet used throughout the Cedar Waxwings (Bombycilla cedrorum)initially period of captivity. Willson and Comet (1993) showed a clear preference for red artificial fruits found that groupsof nestling Northwestern Crows over blue, green, and yellow ones, but fed more hand-raised on single-colored diets (either red, equally on all colors as the testsprogressed (Mc- yellow or “neutral”) did not necessarily select Pherson 1988). Evidence for an overall color the fruit color on which they were raised. This, preference when offered red, blue, yellow and together with our findings, suggeststhat experi- black artificial fruits was not strong among Gray ence with, or exposure to particular food colors Catbirds (Dumetella carolinensis), Swainson’s may not strongly affect fruit color preferences.In Thrushes (Catharus ustulatus) or Hermit our study, exposureto a particular colored main- Thrushes (C. guttutus) (Willson et al. 1990). tenance diet may not have lasted long enough to Individual variation in color preferencesamong affect fruit color preferences. However, longer Silvereyes was greater with the artificial fruits exposure times for crows (Willson and Comet than with R. parabolica fruits, for which all birds 1993) did not alter color preferencesconsistently. preferred red. If artificial fruits are “sub-opti- Furthermore, if Silvereyes made no association mal” stimuli, individual differencesin the pref- between maintenance diet color and test fruit erences of birds might be more apparent than color becauseof the disparity in thesefoods, then with the fruits of R. parabolica. In fact, the lower exposurewould understandably have had no ef- consumption rates of artificial fruits (Fig. 2, a vs. fect on fruit choices.It may be necessaryto ex- b) is consistent with this explanation; different posethe birds to actual fruits of a particular color tastes, sizes, texture and nutritional rewards of prior to testing for any effectof exposureon color the two fruit types could explain the disparity. preferencesamong such fruits. Similarly, differencesin familiarity of Silvereyes It seems surprising that generalist frugivores with the two fruit types may have affected the like Silvereyesshould show such strongand con- handling effort required to consume the fruits. sistent preferences for red fruits in the aviary, Nevertheless,the overall color preferencesof Sil- when they feed on fruits of many colors in the vereyes when presentedwith artificial fruits gen- field (Forde 1986, French 1990). Indeed, the rel- erally reflected those shown between real fruits ative removal rates of the different fruit color in these experiments. morphs of R. parabolica in the field, where Sil- 788 HELEN L. PUCKEY ET AL.

vereyeswere the major frugivore, were not mark- choice by using different pigment systems.That edly different (Willson and O’Dowd 1989). The the overall color preferencesfor artificial and real apparent discrepancy between the fruit color fruits were similar suggeststhat pigment flavors morphs preferred by Silvereyes in the aviary and were unlikely to explain fruit-color preferences in the field suggeststhat either multiple factors in our experiments. affectcolor selectionof R. parabolica or that more Wheelwright and Janson (1985) suggestedthat rigorous field testsare required. Among the fruit studies of fruit color within plant genera may colors we examined, red may be preferred be- yield clearer ecological correlates than investi- cause it is highly conspicuous, i.e., contrasts gations involving unrelated taxa. Genetic fruit- strongly with background color. In the Co- color polymorphisms may provide an even more prosma, reddish fruit colors show more contrast powerful experimental tool because other fruit againstgreen leaves than other fruit-color groups traits are more likely to be similar among color (Lee et al. 1994). In nature, color may serve as morphs. These polymorphisms are widespread, a conspicuous cue or “orienting stimulus” that occurring in a variety of plant speciesin many guidesseed dispersers to a valuable food resource different habitats (Willson 1986, Lee et al. 1988, (McPherson 1988, Willson and Whelan 1990, Willson and O’Dowd 1989). Unfortunately, to Willson 1994). date only a few aviary studieshave yet examined avian color preferences among polymorphic BASIS OF COLOR SELECTION fruits, and, except for our study, none has shown Silvereyes selected fruit colors on the basis of consistent patterns of selection among color hue rather than brightness(Experiment 4). Birds morphs (Willson and Comet 1993, Willson 1994). in general are thought to have well-developed Interestingly, the preferenceof captive Silvereyes color vision, especiallywithin the orange-redpart for the red fruits of R. parabolica in our study of the spectrum (Burkhardt 1982), and a few ex- paralleled their high frequency in the field rela- periments have demonstrated color vision in pi- tive to white and yellow color morphs (Willson geons and domestic chickens (Kare and Rogers and O’Dowd 1989). 1976). To our knowledge, this is the first study Our results support the hypothesis that red to provide evidence that fruit-color choiceby any fruit coloration is an adaptation for promoting frugivorous bird is based on hue. The results of avian frugivory and seed dispersal (Willson and surveys of the colors of bird-dispersed fruits and Whelan 1990). Although this study examined of many experiments on fruit color choice in preferences of only one of fiugivorous birds have been assumed on the basis of hue bird feeding on just two fruit types, the strong alone. If Silvereyes had selected fruits on the preference for red was consistent with the high basis of brightness in Experiment 4, some im- frequency and conspicuousnessof red fruits re- portant inferences about color preference drawn ported in nature. in previous studies might have required re-eval- ACKNOWLEDGMENTS uation. Certainly, more studies on a variety of frugivorous speciesneed to be carried out, be- Mary Willson inspired us to use the fruit-color poly- cause it is unlikely that all frugivorous birds se- morphism of Rhagodiaparabolica to investigatecolor preferences.J. M. Cullen provided expert help with the lect fruit color on the same basis. Although Sil- statisticalanalyses and writing ofprograms, and shared vereyes selectedfruits primarily on the basis of his broad knowledaeofbird behavior. P. McCloud and hue, some individuals appeared capable of rec- M. Bailey of the Dipartment of Mathematics, Monash ognizing differencesin brightnessand responding University, guided us in using GLIM. Lindsay and Sonia Stevensallowed us to collect Silvereyeson their to them. property.Peter Fell and Graeme Farringtonhelped with Pigments are not necessarily without flavor, captureand maintenanceof birds. W. R. A. Muntz and and different tastes associatedwith them might W. S. Jaggeradvised us on measuringbrightness levels account for the observedcolor preferencesof Sil- of artificial fruits. David, Susan and Tim Puckey as- vereyes. We could not distinguish differencesin sistedwith aviary experiments.J. M. Cullen, W. R. A. Muntz and M. F. Willson improved the manuscript. flavor among artificial fruits of different colors or among the polymorphic fruits of R. parabol- LITERATURE CITED ica, although the birds may have been able to do AITKIN,M., D. ANDERSON, B. FRANCIS,AND J. HINDE. so. We attempted to evaluate the possibility that 1987. Statisticalmodelling in GLIM. Clarendon taste was a confounding variable in fruit color Press,Oxford. FRUIT COLOR CHOICE BY SILVEREYES 789

BURKHARM,D. 1982. Birds, berries and UV. A note MOE~MOND,T. C., J. S. DENSLOW, D. L. LEVEY,AND on some consequencesof UV vision in birds. Na- E. SANTANA. 1987. The influence of context on turwissenschaften69: 153-157. choicebehaviour: fruit selectionbv trodcal birds, CRAWLEY,M. J. 1993. GLIM for ecologists.Blackwell p. 229-254. In M. L. COMMONS, A. kACELNIK; Scientific, Oxford. & S.J. Srrm-r~~wo~r~ [eds.],Quantitative anal- FORDE. N. 1986. Relationshios between birds and vses of behaviour. Lawrence Erlbaum, London. fruits in temperate Australia, p. 42-58. In H. A. MO&EN-MOORE, A. L., AND M. F. WIL&N. 1982. Forde and D. C. Paton [eds.], The dynamic part- On the ecologicalsignificance of fruit color in Pru- nership:birds and plants in southernAustralia. D. nus serotinaand Rubus occidentalis:field experi- ;iaWoolman, Government Printer, South Austra- ments. Can. J. Bot. 60: 1554-1560. MUNTZ, W. R. A. 1974. Comparative aspectsin be- FRENCH,K. 1990. Evidence for frugivory by birds in havioral studiesof vertebrate vision, p. 155-226. montane and lowland forest in South-eastAustra- In H. Davson and L. T. Graham, Jr. teds.], The lia. Emu 90: 185-189. eye, Vol. 6. Academic Press, New York. GAUTIER-HION,A., J. M. DUPLANTIER,R. Qums, F. MYERS,B. A., D. H. ASHTON,AND J. A. OSBORNE. FEER,C. SOURD,J. P. DEVOUT, G. DUBOST,L. 1986. The ecology of the mallee outlier of Eu- EMMONS, C. ERARD, P. HECKETSWEILER,A. calyptusbehriana F. Muell near Melton, Victoria. MOUNGZI,C. ROUSSILHON,AND J. M. THIOLLAY. Aust. J. Bot. 34: 15-39. 1985. Fruit characteristicsas a basisof fruit choice RIDLEY,H. N. 1930. The dispersalof plants through- and seed dispersal in a tropical forest vertebrate out the world. L. Reeve, Ashford, Kent. community. Oecologia 65: 324-337. TURCEK,F. J. 1963. Color preferencesin fruit- and GOLDSTEIN, E. B. 1989. Seeing colors, p. 111-137. seed-eatingbirds. Proc. Int. Omithol. Congr. 13: In E. B. Goldstein [ed.], Sensationandperception. 285-292. - 3rd ed. Wadsworth Publishing, Belmont, CA. WHEELWRIGHT,N. T., ANDC. H. JANSON.1985. Col- KARE,M. R., ANDJ. G. ROGERS.1976. Senseorgans, ors of fruit displaysof bird-dispersedplants in two b. 29-52. In P. D. Sturkie [ed.], Avian physioiogy: tropical forests. Am. Nat. 126: 177-799. 3rd ed. Springer-Verlag,New York. WHELAN. C. J.. AND M. F. WILLSON. 1994. Fruit KNIGHT, R. S., AND W. R. SIEGFRIED.1983. Inter- choice in-migrating North American birds: field relationshipsbetween type, sizeand colourof fruits and aviary experiments. Oikos 7 1: 137-l 5 1. and dispersalin SouthernAfrican trees. Oecologia WILKINSON.L. 1990. SYGRAPH: the svstem for 56: 405412. graphics.SYSTAT, Inc., Evanston, IL. ’ KORNERUP,A., ANDJ. H. WANSCHER.196 1. Methuen WILLSON,M. F. 1986. Avian frugivory and seeddis- handbookof colour. 3rd ed. Politikens Forlag, Co- persal in eastern North America. Current Omi- penhagen. thol. 3: 223-219. LEE, w. G., I. L. WEATHERALL, AND J. B. WILSON. Wnrso~, M. F. 1994. Fruit choicesby captive Amer- 1994. Fruit conspicuousnessin some New Zea- ican Robins. Condor 96: 494-502. land Coprosma (Rubiaceae) species. Oikos 69: WILLSON,M. F., ANDT. COMET. 1993. Food choices 87-94 by northwesterncrows: experiments with captive, LEE,W. G., J. B. WILSON,AND P. N. JOHNSON.1988. free-ranging and hand-raised birds. Condor 95: Fruit color in relation to the ecology and habitat 596-615. of Coprosma(Rubiaceae) species in . WILLSON,M. F., D. A. GRAFF, AND C. J. WHELAN. Oikos 53: 325-33 1. 1990. Color preferencesof frugivorous birds in LEON, D. J., AND A. GRAJAL. 1991. Evolutionary relation to the colors of fleshv fruits. Condor 92: implications of fruit-processinglimitations in Ce- 545-555. da; Waxwings. Am. Nat. 138:-171-189. WILLSON, M. F., A. K. IRVINE, AND N. G. WALSH. MCPHERSON.J. M. 1987. A field studv ofwinter fruit 1989. Vertebrate dispersal svndromes in some preferencesof Cedar Waxwings. Condor 89: 293- Australian and New Laland plant communities, 306. with geographiccomparisons. Biotropica 2 1: 133- MCPHERSON,J. M. 1988. Preferencesof Cedar Wax- 147. wings in the laboratory for fruit species,color and WILLSON,M. F., ANDD. J. 0’Dow~. 1989. Fruit color size: a comparisonwith field observations.Anim. polymorphism in a bird-dispersed shrub (Rha- Behav. 36: 961-969. godia parabolica) in Australia. Evol. Ecol. 3: 40- MOERMOND, T. C., J. S. DENSLOW,D. J. LEVEY,AND 50. E. SANTANA.1986. The influence of morphology WILLSON,M. F., ANDC. J. WHELAN. 1990. The evo- on fruit choice in Neotropical birds, p. 137-146. lution of fruit color in fleshy fruited plants. Am. In A. Estradaand T. H. Fleming [eds.],Frugivores Nat. 136: 790-809. and seed dispersal. W. Junk Publishers, Dor- drecht. 790 HELEN L. PUCKEY ET AL.

APPENDIX 1. Selectedcharacteristics of artificial fruits and polychromatic fruits of Rhugodiaparabolica used in all preference trials. For R. parabolica, the mean is followed by SE in parentheses(data from Willson and G’Dowd 1989). SE is not given for artificial fruits as they were made to the same recipe (exceptfor the pigments) and in the same fruit molds.

Artificial fruits Rhagodia parabolica Trait Red Y&XV White Red Y&XV White

Pigments azorubine tartrazine titanium betacyanins betaxanthins flavonoids tartrazine ponceau-4R dioxide Diameter (mm) 6.5 3.2 (0.05) 3.2 (0.07) 3.4 (0.06) Sugars(% dry mass) Fructose 32.6 21.3 (0.6) 25.5 (0.5) 23.3 (0.5) Glucose 32.6 23.5 (0.4) 28.7 (0.4) 22.8 (1.3) Sucrose 0 7.3 (0.5) 4.7 (0.3) 5.3 (0.3) Total 65.2 53.7 (1.6) 58.1 (0.4) 49.4 (1.9) Gelatin (O/adry mass) 34.8 - - Water (O/o) 81.3 79 (0.3) 17 0.6) 76 (0.6)