THE CONDOR -- 1990 A JOURNAL OF AVIAN BIOLOGY SW17

Volume 92 Number 3

The Condor92545-555 0 The CooperOrnithological Society 1990

COLOR PREFERENCES OF FRUGIVOROUS BIRDS IN RELATION TO THE OF FLESHY FRUITS’

MARY F. WILLSON,* DANIEL A. GRAFF,AND CHRISTOPHERJ. WHELAN~ Department of Ecology,Ethology, and Evolution, Universityof Illinois, SherfordVivarium, 606 E. Healey Street, Champaign, IL 61820

Abstract. Aviary experimentsdemonstrated that threeNorth Americanfmgivorous mi- grantbird speciesoften exhibit colorpreferences, when factors such as taste,nutrition, and accessibilityto food sourcesare equal. Individual birds differed in initial preferences, transitivity, and temporal stability of color preferences.There was little tendency of these birds to favor and , which are the most common colors of fruits of bird-dispersed plants, and a weak tendency to reject . We suggestthat avian color preferencesmay not provide strong selection favoring the evolution of the common fruit colors, and that the frequency distribution of fruit is best explained in other ways. Key words: Color preferences;fmgivory; fruit color; Gray Catbird; Swainsons’ Thrush; Hermit Thrush;Dumetella carolinensis: Catharus ustulatus; C. guttatus.

INTRODUCTION proposed,and several factors may have contrib- Fleshy fruits consumed by avian frugivores are uted to this distribution (Willson and Whelan, often colored red or black, sometimes , pur- in press). To begin to assessthe importance of ple, , or , but rarely , yellow, different factors that may help account for the , or (Janson 1983, Willson et al. observed distribution of fruit colors, we examine 1989). Fruit colors are commonly considered to the simple hypothesis that the frequency distri- increase the conspicuousnessof a ripe fruit crop bution of colors reflects the color preferencesof and/or attract the birds that eat fruits and dis- avian frugivores. Under this hypothesis, one perse the enclosed seeds (Darwin 1859, Ridley would expect that red and black would be pre- 1930, van der Pijl 1982). However, there is little ferred most frequently and consistently, followed documentation of the effects of fruit colors on by blue and white, and then by yellow, green, avian foraging decisions (McPherson 1988) or and brown. on variation in color preferenceseither within or Here we examine the following questionsabout among fiugivorous bird species. the attractivenessof different colors, as measured An explanation of the frequency distribution by color preferences of some frugivorous birds of fruit colors must account for the predomi- of the deciduous forest region of eastern North nance of red and black, the lesser occurrence of America: (1) What colors (of food) are preferred blue and white, and the rarity of yellow, green, (or avoided) by fruit-eating birds? (2) How vari- and brown. A number of hypotheseshave been able are color preferences among conspecific birds? Do different bird specieshave similar col- or preferences?(3) Does a given bird make “tran- ’ ’ Received2 November 1989. Final acceptance26 sitive” (i.e., internally consistent, see Methods) February1990. choices?(4) Are color preferences stable in the * Presentaddress: Forestry Sciences Laboratory, P.O. Box 20909, Juneau,AK 99802-0909. face of exposure to other food colors? (5) Does 3Present address: The Morton Arboretum, Lisle, IL the background against which the fruits are dis- 60532. played affect color preferences?

15451 546 M. F. WILLSON, D. A. GRAFF AND C. J. WHELAN

METHODS , red, or blue. All of these colors are ren- dered as perceived by human eyes; we do not SUBJECTS know what the birds actually see. Under the ini- We used Gray Catbirds (Dumetella carolinensis), tial hypothesis, the expected outcome would be Swainson’s Thrushes (Catharus ustulatus), and a preference for red and black, followed by blue, Hermit Thrushes (C. guttatus) for these experi- and then yellow. ments. The thrushes were entirely passagemi- Because we were interested in whether color grants, but catbirds probably included both mi- preferences are transitive (e.g., if red > black, grants and summer residents. The birds were and black > blue, then red > blue, see Moer- captured locally in mist nets in the fall of 1988 mond and Denslow 1983) we used a pairwise and housed individually in six aviaries (2 x 2 design, in which the direction of preference for x 2.5 m) at the Shelford Vivarium, University each pair of colors was determined indepen- of Illinois. Birds were held lessthan 3 weeks and dently, rather than a four-way “cafeteria” design. were releasedat the end of the experiment; three The order of the pairwise comparisons was de- birds escaped during the trials. The number of termined from a random-number table, with the birds tested was constrained by the number of constraint that alternating pairwise comparisons aviaries available, the need to release birds be- on the same day used all four colors (e.g., red vs. fore their local seasonof migration was over, and blue, alternating with black vs. yellow). The first the number escaping. Captive birds were fed comparison alternated with the second, until a chiefly on a banana-based maintenance diet de- predetermined number of replicates (see below) signed for fruit-eating birds (Denslow et al. 1987) was complete. Likewise, the third comparison but augmented with eight teaspoons of glucose alternated with the fourth, and the fifth alter- (per 2 l), supplemented with mealworms (Te- nated with the sixth. nebrio molitor), wax worms (Galleria mellonel- Because catbirds reacted quite differently in la), and real fruit, of various speciesand colors, the aviaries from both thrush species,we used collectedlocally. All birds were given several days slightly different experimental protocols for the to become accustomed to aviary conditions be- different species.Catbirds adjusted quickly to the fore testing..Although some of the test birds may aviaries and readily foraged in our presence. In have hatched in 1988, all were fully fledged and a given foraging trial, two cubes (of two colors) capable of foraging independently. Two cohorts were placed in each of four separatepetri dishes of catbirds were used: Cohort 1 (birds l-6), 8- (30 mm diameter) arrayed along a wooden perch. 12 September; Cohort 2 (birds 7-12), 28 Sep- Each bird was allowed to feed from this display tember-l 1 October. Swainson’s Thrushes (n = until four of the eight cubes had been eaten (usu- 4) were tested on 19-23 September, and Hermit ally < 15 min). For each catbird, there were four Thrushes (n = 6) on 16 October-2 November. replicates of each trial; we did not attempt to All birds used in these experiments ate well in analyze any possiblepatterns of variation among captivity. replicates. Because equal numbers of cubes of each color in each comparison were available, EXPERIMENTAL PROCEDURE we defined no color preference as no statistical Initial color preferences.Experiments consisted difference (determined with the binomial exact of paired choices of maintenance diet (prepared test) in the number of cubes of each color taken. in 4-mm cubes) treated with commercial food Most experiments were conducted between coloring. Colors used were red, blue, yellow, and 07:OOand 14:O0. “black.” An approximate indication of the colors In contrast to the catbirds, neither thrush achieved is given by comparison to Smithe’s species would eat in our presence. For these (1975) color standards: red #l 10, blue #68, yel- species,therefore, we weighed out a known quan- low #18, and black #89. These colors were not tity of dietary cubes in 1O-cm petri dishes (each intended to mimic those of any particular wild color in a separate dish), which were placed in fruits, but they fall within the natural range of the middle of the aviary floors. Placement of fruit colors (to our eyes). Black was produced by dishes with different colors of food was alter- a mixture of roughly equal parts of red and blue, nated to avoid positional bias. We then allowed producing an extremely dark purple; many nat- the birds to feed until roughly half of the total urally black fruits also are, in fact, very dark food was eaten (usually <2 hr) and weighed the FRUIT COLOR PREFERENCES IN FRUGIVORES 547

remainder. The difference was the amount eaten against a background of green foliage on the par- (a small weight loss, <5%, due to evaporation ent plant. We placed each pair of dietary cubes did not affect the analysis). Experiments were on a spray of green artificial foliage attached to conducted between 07:OOand 17:OO.There were the perch. A bird on the perch could reach any three replicatesofeach pairwise comparison, and of the pairs of cubes with two or three hops. the data were analyzed by two-way (bird, color) Results of these trials were compared with those ANOVA (seeSokal and Rohlf 198 1, p. 344-348) described earlier, in which the cubes were pre- and t-tests. The t-tests were done separately for sented in petri dishes. Trials that used different each bird and each pairwise color combination backgroundswere alternated until four replicates (see Appendix 2). Although there was a direc- were completed. A significant shift of preference tional prediction for most pairs of colors (e.g., was indicated by direct comparison of the pref- red > blue), we used a conservative two-tailed erences exhibited on the two different back- test to establish whether there was a significant grounds, using x2 (P < 0.05). difference in the amount eaten. The observed direction of color preference was then deter- RESULTS mined by inspection. This procedure allowed the identification of preferences that were opposite INITIAL COLOR PREFERENCES in direction to the prediction. Catbirds. The evidence of a general preference Stability of colorpreferences. Because most in- for red and black is not strong (Table 1). Of all dividual birds of all three species exhibited 62 trials, 40 (65%) indicated no significant pref- marked preferences for certain colors, we also erence between paired colors. Of 46 trials in- examined the temporal stability of these pref- volving red or black with some other color (i.e., erences. We attempted to extinguish individual blue or yellow), only 16 showeda significant pref- preferences for a certain color by keeping each erencefor red or black. Twenty-five trials showed bird on a maintenance diet of a nonprefetred no preference for red or black over the other color. During continuous exposure to the non- color, and five trials indicated a preference for preferred color, the birds might learn that certain the other color. Thus, 65% (30/46) of the trials characteristics of both colors (e.g., flavor, tex- failed to indicate a preference for red or black. ture) were identical, and thus decreasetheir de- Preferences for blue vs. yellow were also not gree of discrimination between the two colors. marked: Only one of five birds preferred blue Here the question concerned the rate at which (Table 1). Although none of the catbirds discrim- the birds learned to associatethe initially non- inated between red and black, red was favored preferred color with food of equal value to that over blue and yellow more often than black: Four of the preferred color. of 11 birds preferred red to blue, and nine of 12 Catbirds and Hermit Thrushes were used as birds preferred red to yellow, but only three of subjects.An initial paired choice was conducted 11 birds preferred black to yellow (but see Ap- as described above for the thrush experiments at pendix l), and no birds preferred black to blue. the beginning of each day. Subsequently, each However, yellow was rejectedmore often than bird was given only the maintenance diet treated the other colors (13 of 28 trials vs. 4 of 28 for with the nonpreferred color (often supplemented blue, 1 of 34 for red, 4 of 34 for black; all x2 > with various real fruits, mealworms, and wax- 3.84, P < 0.05). Also recall that most birds re- worms) until the next trial. This procedure was jected yellow when paired with red. Thus, the repeated for 6 days. Red and yellow were used results suggestsome tendency to reject yellow, in theseexperiments, becausemost birds showed but little tendency to discriminate among the strong preferencesfor one of these colors. Insta- other colors. bility was indicated by a change in color pref- It may be argued that, if many birds have a erence. weak, statistically insignificant, preference for a Effect of background. For certain pairwise color, the collective effect nevertheless could be comparisons, we examined the effect of back- biologically significant. This possibility was as- ground on color choice by catbirds, because of sessedin two ways. (1) For each pairwise com- the possibility that color preferencesmight differ parison, data for all birds were pooled and a with the background against which the color is single x2 performed on the totals. Red was pre- displayed. Many fruits are naturally displayed ferred over all other colors except black, black 548 M. F. WILLSON, D. A. GRAFF AM) C. J. WHELAN

TABLE 1. Summaryof colorpreferences of Gray Catbirds(details in Appendix 1). Statisticalsignificance as follows:* = P < 0.05, (*) = 0.05 < P c 0.10. Entriesin the tableare the numberof birdsthat madethe indicated choice.For eachpair of colors,the first one is labelled1, the second,2, and thesenumerical labels are usedin the left-handcolumn (for example,the data for 1 > 2 indicatethat no birds preferredblack to blue, two birds preferredblack to yellow,three birds preferredred to blue, etc.).For black vs. yellow,see detailed explanation in Appendix 1.

Color combinations Black vs. blue Black vs. yellow Black vs. red Red vs. blue Red vs. yellow Blue vs. yellow (I vs. 2) (1 vs. 2) (1 YS. 2) (1 vs. 2) (I YS. 2) (1 YS. 2) Cohort 1 No. of birds with the followingpreferences: 1>2 0 2 0 3 3 - 1=2 6 3 6 2 3 - 2>1 0 1 0 1 0 - x2 testof heterogeneity: * * Cohort2 No. of birdswith the followingpreferences: 1>2 0 1 0 1 6 1 1=2 3 4 5 4 0 2>1 3 0 0 0 0 : x2 test of heterogeneity: (*) * *

was preferred only over yellow, and blue was yellow was common, one bird (5) showedno sign preferred to both yellow and black. This proce- at all of such a preference (Appendix 1). Bird 5 dure encounters possible risks of pseudorepli- was also unusual in showing a strong preference cation (Hurlbert 1984) and, although a prefer- for yellow over black, and blue over red. ence for red is indicated, black was not totally Intransitivity of color preferencesoccurred for favored. Thus, even by this possibly question- nine of the 12 catbirds, in at least some of the, able analysis, the hypothesis of a red/black pref- color choices. Birds 6 and 12 made transitive erence is not wholly supported for catbirds. (2) choices, insofar as any preferences were exhib- For all birds that showedsome tendency to prefer ited, and the data for Bird 7 were too incomplete one color (i.e., outcome not equal for both colors), to assesstransitivity. the frequency of preferencefor one color of each Thrushes. Both species of thrush exhibited pair was totalled and tested by a sign test. Eight some color preferences in the majority of trials of 11 birds tended to favor red over blue (not a (in contrast to catbirds). However, red and black significant difference), but all 11 birds showing were preferred over the other colors about as a tendency to favor one color over another fa- often as the other colors were preferred to red vored red over yellow (P < 0.01). Only two of and black (Table 2). In six of 14 trials involving 10 birds tended to prefer black to blue and six red or black vs. the other colors, Swainson’s of 10 birds tended to prefer black to yellow. In Thrushes preferred red or black (and in six trials neither case is the observed frequency of ten- they preferred the other colors), and in 11 of 24 dency significantly different from a null hypoth- trials, Hermit Thrushes preferred red and black esis of equal tendency to favor each color. (and in eight they preferred the other colors). The data are perhaps characterized as much Blue was only sometimes preferred to yellow by their variation as by their congruence.x2 tests (three of four Swainson’s Thrushes, one of six of heterogeneity showed that birds of the first Hermit Thrushes). cohort differed significantly with respect to Swainson’s Thrushes had a marginally signif- choicesbetween red and blue, and between black icant tendency to reject yellow more often than and yellow. Birds of the second cohort differed two of the other colors (seven of 11 trials vs. significantly with respectto blue vs. yellow, black three of 12 for blue and red; x2 = 3.4, 0.05 < P vs. yellow, and, marginally, red vs. blue, and < 0.10). Hermit Thrushes, however, rejectedyel- black vs. blue. Although a preferencefor red over low, black, and blue about equally (six to eight FRUIT COLOR PREFERENCES IN FRUGIVORES 549

TABLE 2. Summary of color preferencesof Swainson’s and Hermit thrushes(details in Appendix 2). Individual variation was significantfor all color combinations. Entries in the table are the number of birds that made the indicated choice. For each pair of colors, the first one is labelled 1, the second, 2, and these numerical labels are used in the left-hand column (see legend for Table 1).

Color combinations Black vs. blue Black vs. yellow Black vs. red Red vs. blue Red vs. yellow Blue vs. yellow (1 vs. 2) (I vs. 2) (1 vs. 2) (I vs. 2) (1 vs. 2) (I YS.2)

Swainson’s Thrushes No. of birds with the following preferences: I>2 1 2 1 1 2 3 1=2 0 0 2 2 0 0 2>1 3 1 1 0 2 1 Hermit Thrushes No. of birds with the following preferences: 1>2 1 2 1 4 4 1 1=2 3 0 3 1 1 2 2>1 2 4 2 1 1 3

of 18 trials) and rejected red marginally less often erencesfor red only on the last day. One catbird than black or blue (three of 18 trials, x2 = 3.3, (Bird 9) switched briefly to yellow and then back 0.05 < P < 0.10). Thus, thrushes showed less to red for the duration of the trials. Bird 12 ini- tendency to reject yellow than the catbirds. tially preferred red, vacillated for a couple of If we pool the data for all conspecificthrushes, days, and returned to a preference for red. we encounter not only the potential for pseu- Hermit Thrushes were somewhat more vari- doreplication but also unequal weightingsfor each able than the catbirds. The preference of Bird 2 bird, reflecting the different total amounts eaten. for red was not extinguished during the course Ignoring suchdifficulties for the moment, no clear of the 6-day trial. Bird 6 showed a relaxation of preferencesfor red and black emerge. Swainson’s red-preference only on the 6th day. Bird 5 Thrushes collectively ate more red than other switched from red to yellow on the 2nd day and colors, but black was equivalent to blue and subsequentlymaintained a preferencefor yellow. somewhat lessthan yellow. Hermit Thrushes col- Both Bird 3 and especially Bird 4 switched pref- lectively ate more yellow than red and more blue erencesmore than once during the 6-day trials. than black. (The sample for both speciesof thrush is too small to allow much information to be EFFECT OF BACKGROUND extracted from a sign test like that done for the Background had little influence on color pref- catbird data.) Thus, for thrushes as well as for erencesof catbirds (Table 4), at least for the color catbirds, the hypothesis of a general preference combinations used in these experiments. In only for red/black hues is not well supported. three paired comparisons was there a significant All interactions between bird and color com- preference, and the background on which the bination were significant (ANOVA, P < 0.05) preference was exhibited differed among birds. for both speciesof thrush, again emphasizing the Two birds showed a significant shift of prefer- prevalence of individual variation in color pref- ence, but in both cases the shift was from an erences.Two Swainson’s Thrushes and two Her- insignificant tendency to choose one color (red mit Thrushes made transitive choices and one or black) to a significant preference for the same Swainson’s Thrush and four Hermit Thrushes color. exhibited some intransitivity (one Swainson’s DISCUSSION Thrush had incomplete data). These frugivorous birds often exhibited marked STABILITY OF COLOR PREFERENCES food-color preferences,when other factors were Both catbirds and Hermit Thrushes varied mark- held constant, and these preferencesappeared to edly in the stability of their color preferences be largely independent of the background (see (Table 3). Two catbirds maintained a consistent also Pank 1976). All three speciesof frugivores preference for red. Bird 11 wavered in its pref- exhibited considerable individual variation in 550 M. F. WILLSON, D. A. GRAFF ANDC. J. WHELAN

TABLE 3. Stabilityof red-yellowcolor preferences of Gray Catbirdsand Hermit Thrushes.Entries in the table are the namesof the preferredcolor (R = red, Y = yellow)on eachday of the trial, as determinedby a choice test. Catbirdswere testedtwice on some days (a, b). Followingeach choicetest each bird was fed on the maintenancediet of its nonpreferredcolor, as determinedby that choicetest. Switches are underlined.

Day I 2 3 4 5 6

Gray Catbirds a b a b a b a b 8 R R R R R R R R R R 9t Y R R R R R R R R R 10 R R R R R R R R R R 11 R R R R R R R R R R&Y 12 R R -R&Y -R -R&Y R R R R R Hermit Thrushes 2 R R R R R R 3 Y Y R R Y Y 4* Y R Y Y R R 5 R Y Y Y Y Y 6 R R R R R R&Y t Bird 9 preferredR in earlierchoice tests (Appendix 1) and switchedat the beginningof thesetrials. * Bird 4 preferredR in earlierchoice tests (see Appendix 2) and switchedat the beginningof thesetrials. color preference.The variability observedamong between the cultivars. American Robins (Turdus birds is probably not an artifact of using artificial migratorius) foraging in Ontario vineyards fa- fruits in the feeding trials. Captive birds also vored a black grape variety over a red one of showed marked variation among individuals in similar sweetnessand a larger but less sweetblue preferencetrials using real fruits (unpubl. observ. one, although the varieties differed somewhat in of MFW and CJW, and R. Jung). Preferencesof how much the foliage concealedthe fruit (Brown some individuals were firmly retained for several 1974) and possibly in other factors. Robins and days, even when food flavor, consistency, and starlings in another vineyard tended to forage nutritional quality were identical (seealso Wink- most extensively on blue rather than white or el 1969), but other individuals had more labile red grapes (Brown 1974). Frugivorous birds in preferences (see also references in Wheelwright Ontario also preferred black to pink or yellow and Janson 1985). The experimental fiugivores sweet-cherry cultivars, but the pink variety was showed little evidence of a general preferencefor reported to ripen slightly later than the black ones red or black (or indeed, for any other color); but (Brown 1974). Polymorphic fruits (red, yellow, catbirds and possibly Swainson’s Thrushes tend- white) of Rhagodia parabolica in Australia were ed to reject yellow more often than other colors. eaten nearly equally by avian foragers (Willson Transitivity of color preference was low, in con- and O’Dowd 1989). trast to a high transitivity of preferencesfor fruit Individually caged robins given a three-way types (Moermond and Denslow 1983). choice between red sour cherries (14% sugar, Other studies of avian fruit-color preferences lower pH) and black (14%) and yellow (18%) are few, and mostly come from cultivated fruits. sweet cherries exhibited much variation among European Starlings (Sturnis vulgaris) in Czecho- birds (Brown 1974): some preferred black to red, slovakia preferred red to white cherries (Prunus while others favored red. All birds initially re- sp.) and blue to white grapes (Vitis sp.) (Feare jected yellow cherries, and most eventually ac- 1984). Eurasian Blackbirds (Turdus merula) in cepted them (after 2-6 days), in some caseseven Bavaria ate both red (ripe) and yellowish (unripe) favoring them. The robins differed markedly in cherries, despite the difference in taste and con- how fast the yellow cherrieswere accepted.Small, sistency, but favored red over white currants captive flocks of starlings often preferred black (Ribes sp.) (Diesselhorst 1972). The taste of the or red cherries initially, but some flocks quickly currants may differ (White 1789 in Snow and (~2 hr) accepted yellow cherries (Brown 1974, Snow 1988) and other factors may also differ Stevens and DeBont 1980). FRUIT COLOR PREFERENCES IN FRUGIVORES 551

McPherson (1988) showed that two captive TABLE 4. Color preferencesof Gray Catbirds when flocks (tested separately but data pooled) of Ce- food was presented on different backgrounds.Signifi- dar Waxwings (Bombycilla cedrorum) collec- cant preferencesare marked *; significantshifts of pref- erencesare marked S. tively ate more red than yellow dietary mash and ate little blue or green. A second trial (about 1 Number eaten week later) was done with the same two flocks Catbird In dishes On leaves but with greater accessto food for all flock mem- Black : Blue Black : Blue bers; in this case the birds collectively favored 1 8:8 I:9 red and blue and, to a somewhat lesser extent, 2 7:9 8:8 yellow, and again ate relatively little green. This 3 7:9 6:lO study was not designed to examine individual 4 8:8 9:7 preferences,and the most notable results are that 6:lO 8:8 the favored colors shifted somewhat between : 11:5 8:8 trials (although red was included in the top rank) Black : Yellow Black : Yellow but that green was consistently rejected. 8 5:ll 917 Many of the field observations of “color pref- 9 9:7 14:2* S 10 6:lO 911 erences” reported above could be confounded by 11 7:9 9:7 other variables, suchas crop size, nutritional val- 12 16:0* 10:6 S ue, palatability, and ripening times. Neverthe- Red : Black Red : Black less, they indicate that foraging choices can be 8 10:6 12:4* variable with respect to fruit color. The experi- 9 8:8 8:8 ments with captive birds document both the 10 917 8:8 variability of color choicesand the ability to learn 11 818 9:7 to eat food that is initially not favored. In ad- 12 6:lO I:9 dition, flower-foraging birds readily learn to visit flowers of many colors (referencesin Willson and Whelan, in press).Although foraging experience fruit types (Levey et al. 1984). Visual sensitivity no doubt influences food-color choices, the nat- may vary among individuals (e.g., Jacobs 1977, ural experience of fiugivores with the common for the squirrel monkey, Saimiri sciureus). Se- red and black native fruits does not seem to pro- lection for lines of different color preferences duce an overwhelming preferencefor fruit of these (Kovach et al. 1976, Kovach 1980) indicates a colors. Even if young birds have any innate pref- genetic basis for at least some of these kinds of erence for thesecolors, by the time they are fully variation. However, color preferencesalso may grown they often seem to have acquired prefer- depend, in part, on the diet of the individual, ences for colors other than red or black. specifically the inclusion of carotenoids, which A remaining possibility is that experiments us- are important in (Kovach et al. 1976). ing some other , blue, and yellow Furthermore, some variation may arise as off- in artificial fruits might yield different results, spring follow their parents and learn to forage especially if a certain shadematched that of some (Werner and Sherry 1987, Peacock and Jenkins particularly favored natural fruit. However, the 1988). It seems reasonable to expect more vari- extensive variation among individuals in pref- ation among young birds that are learning to erencesfor natural fruits observed in both aviary foragethan among older, more experiencedbirds. and field (pers. observ.) suggeststhat the use of A major finding of our study is that color pref- other shades of artificial fruits would also pro- erencesof many individuals were at least partly duce results demonstrating variable preferences. intransitive. In some individuals, preferencesfor Variation among individuals in many aspects some color combinations were transitive, while of foraging has been reported (e.g., Ficken and preferences for other color combinations were Kare 1961, Joem 1988, Ritchie 1988, Rowleyet not. Transitivity of food preferences is one cri- al. 1989, other referencesbelow), so that varia- terion for “rational” decision making (Moer- tion in color preferencesmay not be surprising. mond and Denslow 1983) and has been invoked Individual variation in avian color preferences as support for foraging preferences being based in other contexts is known (e.g., Kear 1966) as on some kind of cost/benefit analysis (e.g., Moer- is variation in strengthsof preferencefor different mond and Denslow 1983, Levey et al. 1984, 552 M. F. WILLSON, D. A. GRAFF ANDC. J. WHELAN

Whelan 1989). Furthermore, many studiesof in- inforcement from additional studies of other dividual differencesin foraging behavior explain species, it could help account for the rarity of thosedifferences in terms of nutritional gain (e.g., yellow and green bird-fruits. Partridge 1976, Werner and Sherry 1987, Harder It remains possible that fruit-color preferences 1988). The observed lack of transitivity of color are contingent upon a variety of associatedtraits preferencesfor food items differing only in color, and conditions (e.g., red could be preferred for in our study, may indicate that color preferences certain fruit shapesor sizes,or in certain habitats, are not based on considerations of energy or nu- or by certain birds), such that they are specific trient utilization. Thus, the basis for color pref- to particular circumstances.Such contingency of erencesmay be fundamentally different from the color responsesis known in insects, depending basis for many other foraging decisions. on activity (e.g., courtship vs. feeding, Swihart Indeed, color is not necessarily a reliable in- and Gordon 197 1) or shapeand size of the object dicator of nutritional value of mature fruits. Fruit (e.g., Hill and Hooper 1984; which may then be colors are not generally correlated with other as- related to activity, Prokopy 1968) although in- pectsof fruit composition in mature fruits, either tensity, contrast, and learning may also be im- among species (Willson and Thompson 1982, portant (Prokopy 1968, Prokopy et al. 1982). Wheelwright and Janson 1985) or within species Contingency of color responsesno doubt occurs (Willson and CYDowd 1989). Fruit colors indi- in vertebrates as well-a relevant instance con- cate the degreeof maturity in a variety of species, cerns the fact that red and black can be warning and birds may use color as a cue to the choosing colors for arthropods that are avian prey, but of more mature fruits (e.g.,Moermond et al. 1986; they may be advertising colors in some positive other referencesin Willson and Whelan, in press). sensefor fruits. Nevertheless, the development of mature fruit An alternative hypothesis is that fruit colors color in cultivated cranberries, grapes,and some are a form of long-distance advertisement for other fruits is related to many environmental avian seed-dispersers(e.g., Kemer 1895, Ridley factors, including temperature, , nitrogen, 1930). This hypothesis has not been tested di- precipitation, CO,, and available sugars(e.g., Hall rectly, so far as we can determine, and circum- and Stark 1972, Kliewer and Weaver 197 1, stantial evidence provided mixed support (Will- Kliewer and Schultz 1973, Kliewer 1977, - son and Whelan, in press). Many additional Schmidt 1980). In addition, preferencesfor ecological factors may have influenced the evo- may be affected by other visible traits such as lution of fruit colors, including costs(both direct brightness (e.g., Kear 1966, Delius and Thomp- and indirect) of pigment production and the pos- son 1970) but brightness is not necessarily cor- sible utility of pigments in defense of the fruit related with fruit ripeness (Sherbume 1972). All against herbivores and pathogens (Wheelwright of theseconsiderations introduce variation in the and Janson 1985; Willson and Whelan, in press). utility of color as a signal on which preferences Any comprehensive theory of fruit-color evo- could be based (see also Willson and Whelan, in lution must encompass a spectrum of selection press). pressures.

IMPLICATIONS FOR THE EVOLUTION OF ACKNOWLEDGMENTS FRUIT COLORS We thank T. Audo and J. Turkot for help in aviary The apparent lack of strong and consistent fruit- maintenance, K. Halupka and D. Niven for mist-net- color preferencesmay contribute to the existing ting, and R. T. Holmes, L. Hogan-Warburg, and D. J. range of fruit colors, but color preferencesseem Levey for constructivecriticism of the manuscript.This unlikely to be a strong factor maintaining the work was funded by a grant from the Whitehall Foun- dation to MFW. observed frequencies of fruit colors in the wild. If the earliest frugivores, back in Paleozoic and Mesozoic times, had color preferences as vari- LITERATURE CITED able and labile as those of today seem to be, even BROWN,R.G.B. 1974. The bird damage problem in the beginning of predominance of red and black southernOntario. Can. Wildl. Serv. Rep. Ser. 27: l-56. as fruit colors is difficult to attribute to color DARWIN, C. 1859. The origin of species. Murray, preferencesof fiugivores. If the tendency to reject London. yellow and green (McPherson 1988) receives re- DEL~JS,J. D., AND G. THOMPSON.1970. Brightness FRUIT COLOR PREFERENCES IN FRUGIVORES 553

dependence of colour preferencesin herring gull of distance between fruits on preference patterns. chicks. Z. Tierpsychol. 27842-849. Ecology 65:844-850. DENSLXIW,J. S., D. J. LEVEY, T. C. MOERMOND, AND MCPHERSON.J. M. 1988. Preferencesof cedar wax- B. C. WENTWORTH.1987. A synthetic diet for wings in the laboratory for fruit species, colour fruit-eating birds. Wilson Bull. 99: 13l-135. and size: a comparison with field observations. Drzss~~~oas~.G. 1972. Beerenund Farbenwahldurch Anim. Behav. 36:96 l-969. Vogel. J. ‘Omithol. 113:448-449. MOERMOND,T. C., AND J. S. DENSLOW. 1983. Fruit FEARE,CHRISTOPIW. 1984. The starling.Oxford, New choice in neotropical birds: effects of fruit types York. and accessibilitvon selectivitv. J. Anim. Ecol. 52: FICKEN,M. S., AND M. R. KARE. 1961. Individual 407-420. - variation in the ability to taste. Poultry Sci. 40: MOERMOND,T. C., J. S. DENSLOW,D. J. LEVEY,AND 1402. E. SANTANAC. 1986. The influence of mor- GOLDSCHMIDT,E. E. 1980. Pigment changesassoci- phology on fruit choice in neotropical birds, p. ated with fruit maturation and their control, pp. 137-146. In A. Estrada and T. H. Fleming teds.], 207-217. In K. V. Thimann led.1. Senescencein Frugivores and seed dispersal. Junk, Dordrecht. plants. CRC Press, Boca Rat& FL. PANK, L. F. 1976. Effects of seed and background HALL, I. V., ANDR. STARK. 1972. Anthocyanin pro- colorson seedacceptance by birds. J. Wildl. Man- duction in cranberry leaves and fruit, related to age. 401769-774. cool temperatures at a low light intensity. Hot?. PARTRIDGE,L. 1976. Individual differencesin feeding Res. 12:183-186. efficienciesand feedingpreferences of captive great HARDER,L. D. 1988. Choice of individual flowers tits. Anim. Behav. 24:230-240. by bumble bees:interaction of morphology, time, PEACOCK,M. M., AND S. H. JENKINS. 1988. Devel- and energy. Behaviour 104:60-77. opment of food preferences: social learning by HILL, A. R., AND G.H.S. HOOPER. 1984. Attractive- Beldinn’s around sauirrels Suermovhilusbeldinni. nessof variouscolours to Australian tephritid fruit Behav-E&l. Sociobiol. 22:393-399. flies in the field. Entomol. Exp. Appl. 35: 119-l 28. PROKOPY,R. J. 1968. Visual responsesof apple mag- HURLBERT,S. H. 1984. Pseudoreplication and the got flies, Rhagoletispomonella(Diptera: Tephriti- design of ecological field experiments. Ecol. dae):orchard studies. Entomol. Exp. Appl. 11:403- Monogr. 54:187-211. 422. JACOBS,G. H. 1977. Visual sensitivity: Significant PROKOPY,R. J., A. L. AVERILL,S. S. COOLEY, AND C. within-speciesvariations in a nonhuman primate. A. ROITBERG.1982. Associative learning in egg- Science 197:499-500. laying site selectionby apple maggotflies. Science JANSON,C. H. 1983. Adaptation of fruit morphology 218:76-77. to dispersalagents in a neotropical forest. Science RIDLEY,H. N. 1930. The dispersalof plants through- 219:187-189. out the world. Reeve, Ashford, United Kingdom. JOERN, A. 1988. Foragingbehavior and switchingby RITCHIE,M. E. 1988. Individual variation in the abil- the grasshoppersparrow Ammodramus savanna- ity of Columbian ground squirrelsto selectan op- rum searchingfor multiple prey in a heterogeneous timal diet. Evol. Ecol. 2:232-252. environment. Am. Midl. Nat. 119:225-234. ROWLEY,I., E. RUSSELL,AND M. PALMER. 1989. The KEAR, J. 1966. The pecking responseof young coots food preferencesof cockatoos:an aviary experi- Fulica atra and moorhens Gallinula chloropus.Ibis ment. Aust. Wildl. Res. 16:19-32. 108:118-122. SHERBURNE,J. 1972. Effects of seasonalchanges in KEENER,A. 1895. The natural history of plants. (En- the abundance and chemistry of the fleshy fruits glish translation by F. W. Oliver.) 2 ~01s. Holt, of northeastern woody shrubs on patterns of ex- New York. ploitation by fmgivorous birds. Ph.D. diss., Cor- KLIEWER,W. M. 1977. Influence of temperature, so- nell Univ., Ithaca, NY. lar radiation and nitrogen on coloration and com- SMITHE,F. B. 1975. Naturalist’s color guide. Amer- position of emperor grapes.Am. J. Enol. Vitic. 28: ican Museum of Natural History, New York. 96-103. SNOW,B., AND D. SNOW. 1988. Birds and berries. KLIEWER,W. M., AND H. B. SCHULTZ. 1973. Effect Poyser, Calton, United Kingdom. of sprinkler cooling of grapevineson fruit growth SOKAL,R. R., ANDF. J. ROHLF. 1981. Biometry. 2nd and composition. Am. J. Enol. Vitic. 24: 17-26. ed. W. H. Freeman and Co., San Francisco. KLIEWER,W. M., ANDR. J. WEAVER. 1971. Effect of STEVENS. J.. AND A. F. DEBONT. 1980. Choice bv crop level and leaf area on growth, composition, starlings (Sturnus v. vulgarisL.) among different and coloration of “Tokay” grapes. Am. J. Enol. cherry cultivars. Agricultura (Heverlee) 28:421- Vitic. 22:172-177. 436. KOVACH,J. K. 1980. Mendelian units of inheritance SWIHART,S. L., ANDW. C. GORDON. 1971. Red pho- control color preferencesin quail chicks(Coturnix toreceptor in butterflies. Nature 23 1:126-127. coturnixjaponica). Science 207~549-551. VAN DERPUL. L. 1982. The nrincinles of disnersalin KOVACH,J. K., G. WILSON,AND T. O’CONNOR. 1976. higher plants. 3rd ed. SphngerLVerlag,Berlin. On the retinal mediation of genetic influences in WERNER,T. K., ANDT. W. SHERRY. 1987. Behavioral color preferences of Japanese quail. J. Comp. feedingspecialization in Pinaroloxiasinornata, the Physiol. Psychol. 90: 1144-l 151. “Darwin Finch” of CocosIsland, CostaRica. PNAS LEVEY, D. J., T. C. MOERMOND, AND J. S. DENSLQW. 84:5506-5510. 1984. Fruit choice in neotropical birds: the effect WHEELWRIGHT,N. T., ANDC. H. JANSON.1985. Colors 554 M. F. WILLSON. D. A. GRAFF AND C. J. WHELAN

of fruit displays of bird-dispersed plants in two WILLSON, M. F., A. K. IRVINE, AND N. G. WALSH. tropical forests. Am. Nat. 126:777-799. 1989. Vertebrate dispersal syndromes in some WHELAN,C. J. 1989. Avian foliage structureprefer- Australian and New Zealand plant communities, ences for foraging and the effect of prey biomass. with geographiccomparisons. Biotropica 2 1:133- Anim. Behav. 38:839-846. 147. WILLSON,M. F., ANDD. J. 0’Dow~. 1989. Fruit color WILLSON,M. F., AND C. J. WHELAN. In press. The polymorphism in a bird-dispersed shrub (Rha- evolution of fruit color in fleshy-fruited plants. go& parabolica) in Australia. Evol. Ecol. 2:40- Am. Nat. 50. W~NKEL,W. 1969. Experimentelle Untersuchen an WILLSON,M. F., AND J. N. THOMPSON.1982. Phe- Zuckervijgeln (Coerebidae) in Funktionkreis der nologyand ecologyof color in bird-dispersedfruits, Nahrungssuche:Uber die Bedeutungvon Farben, or why some fruits are red when they are “green.” Formen und Zuckerkonzentrationen. Z. Tierpsy- Can. J. Bot. 60:701-713. chol. 261573-628.

APPENDIX 1. Numbers of dietary cubes of different colors eaten by Gray Catbirds. Statistically significant preferencesare marked *. For Bird 2t, the outcome of the choice between black and yellow could have been 12:4; becauseof this uncertainty, the summary in Table 1 could be that three birds preferred black and two showed no preference. In the rows for each cohort, between-individual heterogeneityis indicated by ** when P < 0.05 and as (**) when 0.05 < P < 0.10.

Color combination Black Blue Black YdlOW Black Red Red Blue Red Yellow Blue YdOW

Bird Cohort 1 ** ** 1 8 8 8 10 12 4* 14 2 7 9 11 !1_ 1: 6 11 5 14 3 7 9 11 5 9 7 12 4* 10 6 - - 4 14 2* 5 11 12 4* 11 5 - - 5 1 15* 9 7 13* 8 - - 6 11 5 14 2* 7 9 1: 5 1: 3* - - ** ** Cohort 2 (**) (**) 7 4 12* - - 16 0* 8 2 14* 11 ; lo 13 <* 13 ; lo 9 3 13* ; 7 8 8 5 11 13 :: 16 0* 10 5 11 6 10 7 9 10 6 16 0* 11 5 11 10 6 7 9 8 8 7 9 16 5 12 5 11 16 0* 10 6 9 7 16 5