TheCondor95:395-400 0 The Cooper Omthological Soaety 1993

AN EXPERIMENTAL TEST OF THE CONTRASTING- HYPOTHESIS OF -BAND EFFECTS IN RED-WINGED BLACKBIRDS ’

KAREN J. METZ AND PATRICK J. WEATHERHEAD Department of Biology, Carleton University,Ottawa, Ontario KlS 5B6, Canada

Abstract. We performed an experiment to investigate effects of red leg bands on the behavior of male Red-winged Blackbirds(Agelaiusphoeniceus). We presentedmodels having either , , or red leg bands to territorial male Red-winged Blackbirds to test the hypothesisthat any color of band that contrastswith the color of the legsmakes an individual appear abnormal or unhealthy and therefore subject to attack. Overall, territorial males respondedequally aggressivelyto the black- and blue-banded models, but spent more time at distancesgreater than 10 m, displayedat lower intensities, and took longer to attack when the model was given red bands. Thus, red bands appearedto make the model initially more threatening to territorial male Red-winged Blackbirds. These results do not support the contrasting-colorhypothesis and suggestthat the effect of red bands is attributable to the bands matching the color of the male’s epaulets. The failure of the contrasting-colorhy- pothesisalso leaves unresolvedthe different outcomesof one previous experiment showing a negative effect of red bands, and two analysesof long-term banding studiesthat detected no effect of red bands. Key words: Red-wingedBlackbird; Agelaius phoeniceus;epaulets; colored bands.

INTRODUCTION than red-banded males with smaller epaulets We recently demonstrated that red leg bands se- (Metz and Weatherhead 199 1). The causeof the verely reduced the ability of male Red-winged territory loss appeared to be increased and more Blackbirds(Ageluiusphocnicrus) to maintain their persistent intrusion by other males, particularly territories, unlike black bands which had no no- territorial neighbors. Thus, we proposedthat red ticeable effect on male behavior (Metz and bands acted as secondary sexual traits in con- Weatherhead 199 1). That study added to grow- junction with the males’ red epaulets, the com- ing evidence that colored leg bands can affectbird bined effect of which was to make red-banded behavior (Burley 1981, 1985, 1986a, 1986b; males appear to be signaling more aggressively Burley et al. 1982; Brodsky 1988; Hagan and than their competitive ability warranted. The in- Reed 1988; Holder 1990). Because colored leg consistency between signal and ability caused bands are widely used in ornithological research, thesemales to be challenged.This effectwas most it is important that we understand the circum- pronounced in males with large epaulets and was stancesthat causecolor bands to affect behavior. most noticeable to their neighbors. Color-band In addition to the practical concerns, under- effects reported by Brodsky (1988) Hagan and standing color-band effects may also shed Reed (1988) and Holder (1990) also involve col- on questions concerning the evolution of colored ors that matched secondary sexual traits, and in secondary sexual traits in birds (Burley 1986b). the last two studies, the effect of those bands was Here, we present an experiment that tests one of also negative. the hypothesesproposed to explain the effect of We also proposedother hypothesesthat might red bands on male Red-winged Blackbirds. explain the negative effect of red bands on male In addition to the general result that red leg Red-winged Blackbirds (Metz and Weatherhead bands caused some males to lose territories, we 1991). Because epaulets are coverable (Hansen also observed that red-banded males with larger and Rohwer 1986) in Red-winged Blackbirds epaulets were more likely to lose their territories (whereas bands are not), it may have been the exposure of an aggressivesignal under inappro- priate circumstances rather than the size of the I Received8 December1992. Accepted 21 Decem- signal per se that made red bands detrimental. ber 1992. An experiment in which feather clipping caused 13951 396 KAREN J. METZ AND PATRICK J. WEATHERHEAD male Red-winged Blackbirds to exposetheir ep- those models did not have leg bands, previous aulets continuously suggestedthat the ability to experience with a model blackbird could reduce cover epaulets is important for some activities the males’ aggression in subsequent exposures (e.g., trespassing on neighbors’ territories), but (Hansen and Rohwer 1986). Therefore, we an- doesnot causemales to lose their territories (Metz alyzed their responsesseparately from those of and Weatherhead 1992). Thus, the inability to naive males. cover red bands appears insufficient to explain Aggressive territory owners can destroy taxi- why red bands are detrimental. dermic models in a matter of minutes (Rohwer Another hypothesis for the negative effect of 1978, pers. observ.). This can causeserious prob- red bands, and that which we test here, is that lems in keeping the stimulus the same when pre- the bands are detrimental simply because they senting the model to several males in succession. contrast with the color (black) of the legs.A con- Therefore, we designed a model that was virtu- trasting color may make males appear abnormal ally indestructible by a male Red-winged Black- or unhealthy and thus, for some reason (e.g., they bird. The model’s body was made from black are less threatening), provoke an attack. If this cotton material stuffed with cotton balls. We fas- hypothesis is correct, then any color that con- tened a male Red-winged Blackbird wing with trasts with the color of the legs should provoke the epaulet fully exposedto each side of the cot- attack in Red-winged Blackbirds. This hypoth- ton body. A male Red-winged Blackbird head esis could also explain the failure to find any was covered with the same black material as on negative effect of red bands in retrospectiveanal- the body so that only the bill was exposed and yses of two long-term studies of banded popu- was attached to the body. Pins with round black lations (Beletsky and Orians 1989, Weatherhead heads of an appropriate size were used as eyes. et al. 1991). In both studies, all birds received Pairs of male Red-winged Blackbird legs given bands (only some of which were red) that con- six (three per leg) royal blue, black, or red plastic trasted with the color of their legs. Therefore, leg bands could also be fastened with Velcro@ to regardlessof the color of their bands, all birds the body. Thus, the only feature of the model would have been similarly affected,and no effect that varied when presented to owners was the specific to red bands would be expected. We test color of the model’s leg bands. The color of the the contrasting-color hypothesis by quantifying red bands matched the color of the model’s ep- the response of territorial male Red-winged aulets. Red bands and the model’s epaulet color Blackbirds to models with either contrasting (red were scored as Inter-Society Color Council-Na- or blue) or noncontrasting (black) color bands. tional Bureau of Standards (ISCC-NBS) number 36 (Munsell renotation 9.2R 3.9/l 2.1). The color METHODS of the blue bands was scoredas ISCC-NBS num- We conducted this experiment in three cattail ber 182 (Munsell renotation 3.OPB 4.3/6.8). The marshes located within 40 km of the Queen’s blue bands appeared to us to present the same University Biological Station in eastern Ontario contrast to the legs as did the red bands. The from 26 April to 22 May 1990. We presented color of the black bands was ISCC-NBS number territorial males with a male Red-winged Black- 267 (Munsell renotation N 0.8). The black bands bird model given red, blue, or black bands. Ap- matched the color of the legs and thus were dif- proximately half the territorial males were “na- ficult for us to see at any distance. ive,” never having been exposedpreviously to a Model presentations were conducted between model blackbird and not used in other experi- 06:OOto 09:OOhours d.s.t., the time when males ments. The remaining “experienced” males were were most active. We placed the model near the also part of another experiment in which we center of a male Red-winged Blackbird’s terri- clipped either the black feathers that are used by tory while the owner was present and then ob- males to conceal their epaulets, causing the ep- served the response from 30-50 m away for 5 aulets to be permanently exposed( “experienced- min. Male Red-winged Blackbird songswere not clipped”), or an equivalent number of contour played during the model presentations. feathers that did not alter the males’ appearance We measured the following aspectsof the own- (“experienced-controls”). In the course of these ers’ behavior during the model presentations (af- treatments these males had been captured using ter Rohwer 1978): (1) T ls’-the latency in sec- model blackbirds and song playbacks. Although onds to the first time an owner reacted to the BAND-COLOR EFFECTS ON BLACKBIRDS 397

x . x X Xl xtf X 00% x QIX X0 .‘.!qj 0 X .’ o 00 X 0 x . 0 X0 .O 0 l X - l 0 0 0 .l . . x Xk 0 X 0

. 0

0 0

0

-3 -2 -1 2 9 4 FURCTION1 l FIGURE 1. Scatterplotof territorial males’ responsesto the model with red (open circles), black (X s), or blue (closed circles) bands using the males’ discriminant function scoresfor Function 1 (time to attack; proximity to model) and Function 2 (song spread rate and intensity). model (up to 300 set). Reactions included flying respond differently upon subsequent presenta- closer to the model, or directing a song spread tions. We randomly assignedwhether an owner display at the model; (2) T HIT-the latency in was presented with the red-, blue-, or black- seconds to the first time an owner struck the banded model. We usually performed 5-l 2 mod- model (up to 300 set); (3) N HITS-the number el presentations per morning. This normally re- of times the owner struck the model; (4) O-2 m- sulted in all three band being presented the amount of time in secondsan owner spent each morning. Thus, potential variation in be- within two meters of the model; (5) > 10 m-the havioral responsescaused by weather or other amount oftime in secondsan owner spent greater factors were randomized. than 10 meters from the model; (6) Total SS- the total number of songspread displays directed by the owner toward the model; (7) The intensity RESULTS of song spread displays given by the owner that Prior experience did not affect the response of were directed toward the model. We classified males to the model. Thirteen of 3 1 (4 1.9%) ex- the intensity of song spread displays based on perienced males attacked the model and 20 of the body postures of singing males. We consid- 42 (47.6%) naive males attacked the model (x2 ered a display not to be a song spreadwhen sing- = 0.23, df = 1, P = 0.63). There was also no ing males’ wings remained at their sidesand their significant difference between naive and experi- tails were not fanned (SSI). A low intensity song enced males in how vigorously they responded spreadinvolved males slightly arching their wings to the model (t-tests for each variable; all Ps > out from their sides,and holding their tails down 0.05). Among experienced males, clipped and but not fanned (SS2). In a medium song spread control individuals were just as likely to attack the wings were arched out and down and the tail the model; 10 of 15 clipped males attacked the was slightly fanned (SS3). In a high intensity song model, and 8 of 16 control males attacked the spread (SS4), the feathers of the epaulet were model (x2 = 0.88, df = 1, P = 0.35). The behavior ruffled, the wings fully arched out and down and of clipped and control males was also similar the tail completely fanned. when reacting to the model (t-tests for each vari- We presented the model only once to each able; all Ps > 0.05). Since prior experience with owner. This minimized the possibility that own- the model and our clipping treatments did not ers would become familiar with the model and affect how males responded to the model, we 398 KAREN J. METZ AND PATRICK J. WEATHERHEAD

TABLE 1. Discriminant function analysis of behavioral differencesof owners to a conspecificmodel that had red, black, or blue colored bands during 5-min activity budget recordings.Variables are defined in the text.

Red Black Blue Significant variables n = 23 n = 25 n = 25 Entered Variable .z x R Wilks ’ X P atstep

T 1ST 42.57 56.16 37.84 T HIT 274.61 230.52 234.60 0.718 0.034 6 N HIT 1.87 3.92 1.68 O-2 m 105.43 172.32 110.56 0.814 0.028 3 >lOm 77.43 26.08 29.84 0.912 0.039 1 Total SS 14.83 14.32 17.16 0.740 0.026 5 SSl 0.70 0.00 0.08 ss2 0.87 0.12 0.04 0.859 0.032 2 ss3 1.57 1.16 0.76 0.781 0.03 1 4 ss4 7.78 10.16 11.40

Function I Wilks ’ Lambda x ’ = 22.35, P = 0.034. Function 2 Wilks ’ Lambda xz = 8.67, P = 0.12. combined all males in our analysis of their re- strongestto the model with black bands, and with sponse to the red-, black-, or blue-banded model. the fifth variable (total song spreads), the re- To determine whether males reacted differ- sponsewas strongestto the model with blue bands ently to the models we first considered each be- (Table 1). Overall, however, the responseto the havioral variable separately. Seven of 23 males model with red bands was most different. A plot attacked the model with red bands, 14 of 25 of individual discriminant function scoresillus- males attacked the model with black bands, and trates this difference, particularly with regard to 12 of 25 males attacked the model given blue function 1 (time to attack and proximity to the bands. Thus, territorial males were just as likely model) (Fig. 1). to attack the model when it was given red, black, or blue bands (x2 = 3.28, df = 2, P = 0.20). We DISCUSSION used analysis of variance to analyze the remain- The results from our experiment provide stron- ing response variables. We found no significant ger support for the hypothesis that it is the color differences in the responsesto the red-, black-, ofthe red bands that is responsible for their effect or blue-banded model (all Fs < 2.37, all Ps > on males’ behavior, than for the hypothesis that 0.10). We then used discriminant function anal- any color contrasting with the leg will have a ysis to see if we could differentiate the males similar effect. Most of the differences we ob- presented with the red-, black-, or blue-banded served, and those most important in discrimi- model from their response to the model. The nating between treatments, were between red discriminant functions were able to correctly bands and the other two colors. Territory owners classify the responseof owners to red, black, and spent more time farther than 10 m from the red- blue bands with 53.4% accuracyusing songspread banded model, gave more low and intermediate rate and intensities, distance from the model, and intensity song spread displays to the red-banded time to first hit (Wilks’ Lambda Function 1: x2 model, and took longer to attack the model when = 22.35, df = 12, P = 0.03; Function 2: x2 = it had red bands. Thus, red bands appear to make 8.67, df = 5, P = 0.12, Table 1). For the first, the model more threatening to territory owners, second, fourth and sixth variables to enter the consistent with our hypothesis(Metz and Weath- discriminant function, responsesto the model erhead 1991) that red bands enhance the signal with red bands appeared to be different from provided by the epaulets. responsesto the model with either blue or black If larger badges are more threatening to other bands (Table 1). Owners spent more time farther males, why did red-banded males lose their ter- than 10 m from the red-banded model, gave more ritories in our previous study (Metz and Weath- low and intermediate intensity song spreads,and erhead 199 l)? The answer may lie in the differ- took longer to hit the model. For the third vari- ence in the duration of exposure of territorial able to enter the discriminant function (time spent males to males or models with large badges. If <2 m from the model), the response appeared selection maintains the reliability ofbadges, then BAND-COLOR EFFECTS ON BLACKBIRDS 399 a male confronted with a rival with large badges responddifferently to red versusblack bands (i.e., should initially respondcautiously becausea large Metz and Weatherhead 199 l), but not to red badge signalsa dangerousrival. However, if con- versus other contrasting colors (i.e., Beletsky and tinued assessmentof the rival indicates that the Orians 1989, Weatherhead et al. 1991). How- badges are too large for the true quality of the ever, because it appears that the effects we have rival, an escalatedresponse should follow. In our observed in our experimental studies are specif- banding experiment (Metz and Weatherhead ically attributable to red bands, we must consider 199 l), it was the neighbors of red-banded males other explanations for the lack of a red-band that were aggressive.These males had previous effect in the long-term studies. Possible expla- experiencewith the red-banded males before they nations include a threshold to the amount of red received their red bands, as well as a period of necessary to produce an effect, a need for the days to continue assessingthem following band- color of red to match that of the epaulets fairly ing. In our model presentations males only had precisely, or the possibility that red-band effects 5 min to both assessand respond to the models. occurred in the long-term studies but were too Thus, in the model presentations there may have subtle to be detected by the methods used been insufficient time for males to determine that (Weatherhead et al. 1991). Until these hypoth- the red-banded model was not as threatening as esesare explored, we are left with the possibility it appeared. that one of our basic research tools is compro- The hypothesis proposed above to explain the mising the phenomena we study. difference in responsewe observed to red bands in our two studies may also reconcile our model ACKNOWLEDGMENTS presentation results with those of Hansen and We thank the Gile, French, Mathews and McTaggert Rohwer (1986). They demonstrated that models families for use of their marshes. Bill Searcvand Bob that had epaulets that were double their normal Montgomerie for commenting on the manuscript, the size were attacked more vigorously by owners Natural Sciencesand EngineeringResearch Council of than models with normal-sized epaulets. In our Canada and the John K. Cooper Foundation for fi- model presentation study, we only recorded the nancial support, and the Queen’s University Biological Station for logistical support. owners’ behavior for the 5 min immediately fol- lowing placement of the model on the males’ territories. Hansen and Rohwer (1986) started LITERATURE CITED their observations after they had moved to their BELETSKY,L. D., AND G. H. ORIANS. 1989. Red bands observation post (40-60 m from the males’ ter- and Red-winged Blackbirds.Condor 91:993-995. BRODSKY.L. M. 1988. Ornament size influencesmat- ritory) and after the male first flew at the model. ing successin male Rock Ptarmigan. Anim. Be- Then they observed the male’s behavior for 10 hav. 36662-667. min. Thus, males in their study had more time BURLEY,N. 1981. Sex ratio manipulation and selec- to assessthe quality of the model. If correct, this tion for attractiveness.Science 2 11:72l-722. explanation suggeststhat males are able to eval- BURLEY,N. 1985. Leg band color and mortality pat- terns in captive breeding populations of Zebra uate rivals quite quickly, although that may only Finches. Auk 102:647-65 1. be true when the rival is an inanimate model. BURLEY,N. 1986a. Sex-ratio manipulation in color- Our hypothesisalso predicts that with longer ex- banded populations of Zebra Finches. Evolution posure, red-banded males such as those we pre- 40:1191-1206. sented in this study, should be attacked more BURLEY,N. 1986b. Sexualselection for aesthetictraits in specieswith biparental care. Am. Nat. 127:415- vigorously. 445. Our failure to support the contrasting-color BURLEY,N., G. KRANTZBERG, AND P. RADMAN. 1982. hypothesis leaves unresolved the different out- Influence of colour-banding on the conspecific comes of our experimental study (Metz and preferencesof Zebra Finches. Anim. Behav. 30: 444-455. Weatherhead 199 1) and the retrospective anal- HAGAN, J. M., AND J. M. REED. 1988. Red color yses of long term studies of color-banded Red- bands reduce fledging successin Red-cockaded winged Blackbirds (Beletsky and Orians 1989, Woodveckers. Auk 105:498-503. Weatherhead et al. 1991). Had we determined HANSEN,A. J., AND S. ROHWER. 1986. Coverable badgesand resourcedefence in birds. Anim. Be- that the birds responded differently to a conspe- hav. 34~69-76. cific with any color band that contrastswith their HOLDER,K. R. 1990. Ornamental traits in Rock Ptar- black legs, then we could explain why the birds migan: momholoev. behavior and mating success. 400 KAREN J. METZ AND PATRICK J. WEATHERHEAD

M.Sc.thesis, Queen’s University, Kingston, On- ROHWER, S. 1978. Passerinesubadult plumagesand tario, Canada. the deceptive acquisition of resources:test of a METZ, K. J., AND P. J. WEATHERHEAD. 1991. Color critical assumption. Condor 80:73-179. bands function as secondarysexual traits in male WEATHERHEAD, P. J., D. J. HOYSAK, K. J. METZ, AND Red-winged Blackbirds. Behav. Ecol. Sociobiol. C. G. ECKERT. 199 1. A retrospectiveanalysis of 28~23-27. red-band effectson Red-winged Blackbirds. Con- METZ, K. J., AND P. J. WEATHERHEAD. 1992. Seeing dor 93:1013-1016. red: uncovering coverable badgesin Red-winged Blackbirds. Anim. Behav. 43~223-229.