Anim. Behav., 1995, 50, 929-943 Femalechoice, parasite load and male ornamentationin wild turkeys RICHARD BUCHHOLZ Department of Zoology, University of Florida (Received 13 June 1994; initial acceptance 3 August 1994; final acceptance 17 January 1995; hfs. number: A71.%?R) Abstract. Wild turkeys, Meleagris gallopavo, exhibit a particularly striking array of ornamental characters, including a bare, brightly coloured head covered with bumpy caruncles and a distensible snood, a black hair-like beard, tarsal spurs and a large fanned tail. These characters are hypothesized to be maintained by female choice. Good genes models of female choice propose that these ornaments serveas indicators of the bearer’s heritable ability to resist parasites, maximize fat deposition or attain greater age. Aesthetic models propose no such ability of ornaments to indicate these correlates of fitness. These models were tested by (1) determining the male characters assessedby captive females, using artificial and live males in mate-choice trials, and (2) examining the correlates of ornamentation in wild-caught males. Females preferred males with longer snoods and wider skullcaps. These traits were negatively correlated with coccidian oocyst loads in the faecesof wild-caught males. Snood length may also be an indicator of male energy reserves. These results are consistent with good genes models of female choice. Other aspects of male ornamentation in wild turkeys were not explained by inter-sexual selection. The wild turkey, Meleagris gallopavo, is an example of a sexually dimorphic speciesexhibiting an array of extravagant behavioural and morpho- logical characters that serve no obvious function other than to attract mates. Males weigh more than twice as much as females and have iridescent feathers, a large fan-like tail, a long hair-like beard projecting from their chest, tarsal spurs and an unfeathered head and neck. The unfeathered neck is covered with pebble-like red bumps called side caruncles, that may fuse to form horizontal bands of bumpy skin (Fig. 1). At the base of the front of the neck are three or four roughly ellipti- cal outgrowths called frontal caruncles. The skull- cap on the crown of the head has thickened skin that ranges in colour from red to white and light blue and appears to change depending on the motivational state of the animal. Perhaps most striking is a distensible process at the base of the upper mandible called the snood. Finally, a thin dewlap stretches from beneath the lower mandible down to the frontal caruncles. This extravagantly ornamented bird is well suited to testing models of female choice because it is a highly polygynous Correspondence: R. Buchholz,Department of Biology, Northeast Louisiana University, Monroe, LA 71209- 0520, U.S.A. (email: bibuchholz@alpha.nlu.edu). Hypotheses of indirect benefits of female choice fall into two types: good genes and arbitrary preferences. Both types of hypotheses predict that females will exert directional selection for the more exaggerated form of a male trait (e.g. they prefer bigger, brighter or louder forms). Good genes models propose that the character in ques- tion is assessedby females because it is an indi- cator of the heritable vigour of the bearer. Male vigour or viability is a measure of a male’s ability to survive acute or chronic factors that limit fitness. There are two principal types of good genes hypotheses. The first type consists of handi- cap models (Zahavi 1975; Grafen 1990). These propose that ornamentation reduces average male 0003-3472/95/100929+ 15 $12.00/O c 1995The Associationfor the Study of Animal Behaviour 0 1995 The Association for the Study of Animal Behaviour species,and males associate with females only for mating and provide no parental care. Although we cannot be sure by which selective process any one of the turkey’s many ornaments has evolved during the history of this species,the evolutionary history of a character can be inferred by examining the selective pressures maintaining such characters in the present. By testing the mating preferences of captive female wild turkeys and examining the correlates of the preferred types of ornamentation in wild males, I investi- gated how female choice may be maintaining the elaborate ornamentation of male wild turkeys. 930 Animal Behaviour, 50, 4 Figure 1. Unfeathered head ornamentation of mature male wild turkey: a: skullcap; b: relaxed mood; c: dewlap; d: frontal caruncles; e: side caruncles. survival, enabling females to assess any individual male’s ability to survive despite his burdensome ornament. To test the handicap principle of ornamentation, it is necessary to experimentally manipulate the quality of ornamentation of free- living males. Because such a test of the handicap model is beyond the scope of this study, I do not discussthe predictions of this hypothesis further. In the other type of good genes models, females assessmale ornamentation because it specifically indicates a second, lessapparent trait determining male fitness. Hamilton & Zuk (1982) proposed that females assessthe quality of male ornamen- tation because these characters are particularly good indicators of the parasite burden of the bearer. In the parasite-assessmentmodel, females benefit from choosing more ornamented males, because their offspring will inherit the male’s ability to avoid deleterious infection. In similar models, the male’s foraging success (Kodric- Brown 1989) or age (Manning 1985, 1989) is honestly indicated by his ornamentation. Arbitrary preference models, on the other hand, hypothesize that the character reveals nothing of the abilities or heritable vigour of the male but is none the lesspreferred by females. Fisher’s (1958) theory of runaway selection proposed that an initial female preference for a male character that came about by natural selection or genetic drift could become genetically linked to the male trait, resulting in elaboration of the trait beyond its original utility. Similarly, other authors (Kirkpatrick 1987; Enquist & Arak 1993) have suggested that specific attributes of male charac- ters, such as colour, shape or frequency, elaborate over evolutionary time because they more effec- tively stimulate existing biases in the sensory sys- tem of the female. In these models, male offspring of females exercising an arbitrary preference have greater reproductive success as a result of the character they inherit. Because the arbitrary pref- erence model is empirically distinguishable only by the absence of a correlation between the male character and measures of male viability, it proves difficult to test. The arbitrary preference model remains a default or null hypothesis in studies where a female preference is detected, but the adaptive reason for the preference is not explained by the good genes model (Buchholz 1992). The only way to distinguish between good genes and arbitrary models of female choice, when it is not possible to monitor the next generation, is to examine the relationship between variation in ornamentation and variation in male viability. Extensive research on the factors limiting population size in wild turkeys suggests at least two major selective pressures, parasites and weather (Dickson 1992). Population surveys have detected over 100 speciesof viral, bacterial, fun- gal, protozoal, metazoal and arthropod parasites that may limit the fitness of turkeys (Davidson & Wentworth 1992) some of which have a severe impact on host health (Forrester et al. 1974, 1980; Hon et al. 1975, 1978; Atkinson et al. 1988; Forrester 1991). Weather conditions in the north- ern areas of the wild turkey’s range affect sur- vivorship and fecundity. Populations subject to deep snow can suffer widespread starvation and an overwintering mortality approaching 60% (Healy 1992a). Starvation during winter is prob- ably not limited to the northern areas of the wild turkey’s distribution. A 20-fold increase in visits to artificial feeders between January and March suggests that wild turkeys in Florida may find winter food hard to come by as well (Powell 1965). How might females detect males that are superior in their abilities to cope with these factors? Because parasites are a problem for wild turkeys, the Hamilton & Zuk (1982) theory of Buchholz: Female choice in wild turkeys 931 matechoice as an adaptation against parasitism seems a plausible explanation for the maintenance offleshy structures in this species.If turkey hens use male ornamentation to assess parasite burden, theintensity of infection by the most deleterious parasitesshould be inversely correlated with the qualityof male ornamentation that females use to choosea mate. To be favoured by females, these ornaments should be better indicators of parasite burden than other anatomical structures, such as bill or tarsus length. Adult male wild turkeys begin their energetic courtship displays in early spring, relying on their fat reserves for sustenance (Pelham & Dickson 1992). If the ability to access food and store energyfor surviving winter starvation is an impor- tant determinant of survival, as previous research suggests,females may use the quality of a male’s display as an indication of stored energy reserves. If this is the case, female choice should rely heavily on the most energetic parts of the male display and on any other characters that might indicate energy reserves, such as body condition. Age is perhaps the most honest indicator of good genes. Males that survive despite parasites, predators and food shortages have demonstrated their superior ability merely by remaining alive. How likely is it that longevity is a good indicator of fitness in wild turkeys? Annual mortality of male turkeys varies considerably (29967%) depending on habitat quality and hunting pres- sure (which may account for as much as one-third of the mortality; Vanglider 1992). Nevertheless, many males live as long as 5 years, and some attain ages of lo-12 years (Kelly 1975). Males continue to increase their weight as they age (Kelly 1975) and may experience higher fitness becausetheir greater girth allows them to exclude other males from receptive females. Thus females may choose mates based on ornamentation that reliably reveals male age, because their own sons will then be more likely to survive and reproduce. Although beard length and body mass are posi- tively correlated with age, spur length is thought to be the best indicator of yearly age classesin eastern wild turkeys, M g. silvestris (rz0.78; Kelly 1975; Steffen et al. 1990). This correlation suggeststhat females should choose to mate with males with longer spurs. The arbitrary preference models for the evol- ution of male ornamentation predict no relation- ship between ornamentation and male survival or condition. If the characters chosen by females are not related to any of the factors that limit turkey survival or fecundity, it seemsreasonable to con- clude that they are maintained purely by the increased mating successof the male without any direct increase in the reproductive successof the female. I tested these hypotheses for the main- tenance of ornamentation in wild turkeys with experimental data collected from mate-choice trials in captivity and correlational data collected from wild-caught males. GENERAL METHODS Captive Subjects One-day-old wild turkey poults were purchased in May 1991 from a gamefamr (L&L Pheasantry, Hegins, Pennsylvania) whose large breeder flock was interbred with wild males, M. g. silvestris, until a decade ago. Poults were reared indoors under heat lamps with gamebird starter feed (Purina Startena, 30% protein) and water pro- vided ad libitum. After 8 weeks the birds were transferred to an outdoor pen (5.3 x 5.3 x 2.2 m) at the Florida Museum of Natural History’s Special Projects Laboratory. After 14 weeks of age, males were maintained separately from the females and visually isolated from each other in two groups (N, = 16, N2= 17) in sand-floored aviaries (5.3 x 5.3 x 2.6 m). All females were moved to a cement-floored pen (5 x 12 x 4 m) that was visually isolated from the males. Animals were fed daily (Purina Grower or Maintenance, 19 or 12% protein, respectively); they only received as much feed as they could consume in 34 h. Additionally, the birds received a variety of green forage (e.g. cut grass, bamboo leaves) twice a week. Water was available ad libitum. I observed only limited displaying by the cap- tive males during pilot studies in early 1992. I attributed the low occurrence of display to sup- pression by dominant males in the relatively crowded group housing. To alleviate crowding, I housed all males individually in cylindrical wire cages (1.3 m diameter x 1 m high) at 15 months of age. Black plastic dividers prevented males from physically interacting with males in neighbouring pens, although they could see other males in pens 4 m away. Each male received approximately 932 Animal Behaviour, 50, 4 0.5 litres of breeder rations (Purina Layena, 20% protein) daily, along with occasional peanuts, wild bird seed, or carrot or apple slices. Water was provided in 05litre containers and was replaced daily. Mate Choice Experiments and Correlates of Male Ornamentation I asked the following questions to test the assumptions and predictions of the good genes and arbitrary preference models. First, do females mate randomly with respect to male traits, and if not, which traits do they assessin prospective mates? Second, are the characters assessedby females particularly good indicators of the male’s response to factors that decrease their fitness? I tested the mating preferences of female wild turkeys in two experiments. In the live male experiment (I), females received a choice between two naturally displaying males. I com- pared characteristics of males chosen by females to the characteristics of males not chosen by females. The male model experiment (2) gave females a choice between two artificial males that differed only in the amount of head ornamen- tation. The model experiment ensures that the character apparently assessed by females in experiment 1 is not merely the correlate of another, unmeasured character actually under assessmentby females. Do binary mate choice trials adequately reflect the conditions under which sexual selection oper- ates in the wild? The mating strategies of male wild turkeys varies from lek-like assemblages of dyads of males (Watts & Stokes 1971) to male defence polygyny (Williams & Austin 1988). Dur- ing the breeding season, single females are some- times encountered outside of their normal range without other hens or an attendant gobbler (per- sonal observation). Thus it seemsthat females can circumvent attempts by dominant males to main- tain exclusive access to them. In fact, females solicit copulation from subordinate males in some situations (A. Clark, personal communication). Given the opportunity for female choice in this species, I believe that the mate choice trials adequately represent the conditions under which sexual selection operates in free-living populations. EXPERIMENT 1: LIVE MALE TRIALS Methods Males I exposed 33 20-month-old males to an artificial photoperiod of 14:lO h 1ight:dark for 4 weeks prior to and during the mate choice tests. Prior to the mate choice trials, males were classified as ‘displayers’ or ‘non-displayers’ based on whether they strutted during a 0.5-h observation period on each day of the week before the first trials. During the trials no non-displayers became displayers, but the reversedid occur. I used the 11 males classified as displayers in the mate choice trials. I measured seven ornamental characters prior to the first trial. (1) Relaxed snood length was measured from the point of attachment at its base to the tip with a small ruler. I measured the snood again after stretching it by attaching a clip to the tip of the snood and pulling on the clip with a Pesola scale to a tension of 30 g. I chose this tension because it amply stretched the snood without pulling it com- pletely taut. (2) The maximum vertical and hori- zontal diameter of each frontal caruncle, as well as its thickness, was measured with a ruler. I con- verted the radii of the frontal caruncles to an approximation of total caruncle area using the equation for the area of an ellipse (A=ab~). (3) I counted the red, polyp-like projections on the neck called side caruncles. (4) I measured the width of each half of the thickened, whitish skull- cap by placing the ruler in a line from the most posterior point at which the skullcap halves meet along the middle of the cranium to a point dorsal to the middle of the eye. (5) I measured the beard from where the ‘hairs’ leave the skin to its greatest length. (6) I measured tarsal spur length from where the spur enters the scaled skin to the distal tip (Kelly 1975). (7) I measured tarsus length from the articulation of the tarsometatarsus with the tibiotarsus to the third scute of the central phalange of the foot. I used mass divided by tarsus length as an index of male body condition and refer to it as such or merely as male condition. Females I exposed 23 20-month-old female wild turkeys to an artificial 14:lO h L:D photoperiod for 4 weeks prior to and during the mate choice tests in January-March 1993. Buchholz: Female choice in wild turkeys 933 Male 1 Holding area Female Male 2 Fiie 2. Mate choice arena. Female is admitted from a holding area to the choice area where she has a choice of two displaying live males (experiment 1) or two artificial males (experiment 2). The thin vertical line represents the hardware cloth divider separating males and female. Experimental design I subdivided the large cage used for housing the females into three sections to serve as a mate choice arena (Fig. 2). The female flock was housed in one half of the aviary during the experiment. Each female was admitted singly to the choice area through a door in the opaque plastic barrier that prevented the untested females from seeing the displays of the males or the choices made by subject females. An additional opaque barrier prevented the males from directly interacting with one another. The subject female was separated from the males by hardware cloth that had been painted black to make it lessvisible. I tested each female only once and presented it with a unique pairing of males that was never presented to any other female. In total, I used 11 males two to four times each (x=3.3); there was no significant correlation between the number of times each male was presented and the frequency of female choice (Spearman rank correlation: rs=0.51, P=O.14). I randomly assigned males to the two display areas on the afternoon before the trial. I only conducted one trial in any 24-h period from 14 January to 12 March 1993, between 0800 and 1600 hours. During the IO-min pre-trial period before the female was admitted, I recorded the frequency of spontaneous strutting by each male and the degree of snood distention (on a scale of o-4, respectively: not distended, slightly distended, hanging down past the upper mandi- ble, hanging past the lower mandible, hanging past the mid-point of the neck). I recorded snood distention again immediately after the female was admitted and recorded strutting frequency until the female solicited one of the males. Females revealed their choice of mate by ‘crouching’ (Healy 1992b), a conspicuous mating solicitation posture exhibited in front of the chosen male. I also recorded the time that the female spent in front of each male until solicitation. Each trial ended when the female solicited or after 30 min had elapsed. The two males from a failed trial were tested again with a new female the next day. Data analysis I determined the characters associated with female mate choice by comparing the characteris- tics of the male solicited with those of the male not solicited in each trial. Wilcoxon matched-pairs signed-rank tests were used with characters that were independent of other characters but that may not be distributed normally (Siegel 1956). Some of the measures of male ornamentation, however, strongly covaried with each other and with tarsus length (Table I), making it difficult to determine their relative contribution to mate choice. I used principal component factor analysis to isolate independent axes of variation in the correlation matrix of the five most highly covarying charac- ters (tarsus length, spur length, skull cap width, stretched snood length and frontal caruncle area). The inclusion of only five male characters in the factor analysis followed a statistical rule of thumb that the number of variables did not exceed 50% of the sample size (Abacus Concepts 1986). Fron- tal caruncle area was transformed by log (x)“’ when combined with linear measurements in the factor analysis. I subsequently transformed (i.e. rotated) these axes using the varimax solution to generate four orthogonal axes on which the vari- ables were loaded as uniquely as possible (Table II). These axes provided four sets of independent axis scores, each largely composed of only one character. Unlike the strongly covarying raw data, the axis scores could be used to determine which of these five characters independently was most important to a female’s choice of males. For each trial I calculated the differences between the scores of the males presented to the female by subtracting the score of the male not chosen from that of the male that was chosen. If females chose males at random with respect to these characters, the expected distribution of the differences between preferred and not preferred males would have a mean of zero. A positive mean is expected if females choose males with the larger 934 Animal Behaviour, 50, 4 Table I. Pearson correlation matrix of morphological variables and average strut rate of the 11 males used in the mate choice trials Variable Tarsus Mass Spur Beard Side Fc Fc Snood r Snood s car area depth Skullcap Tarsus 1.00 Mass 0.91 1.00 Spur - 0.66 - 0.55 1.00 Beard 0.18 0.32 - 0.13 1 .oo Snood r 0.65 0.54 - 0.47 - 0.46 1 .oo Snood s 0.32 0.22 - 0.45 - 0.48 0.83 1.00 Side car - 0.03 0.01 0.03 - 0.05 0.15 0.10 1 .oo Fc area 0.45 0.45 - 0.50 0.22 0.34 0.17 0.11 1 .oo Fc depth - 0.03 - 0.05 - 0.28 0.06 0.12 0.43 0.10 0.56 1.00 Skullcap - 0.04 0.12 - 0.09 - 0.13 0.45 0.62 0.20 0.12 0.33 1 .oo Strut rate -0.16 0.07 - 0.07 - 0.42 0.29 0.47 0.18 - 0.12 0.09 0.51 Snood r: relaxed snood length; snood s: stretched snood length; side car: number of side caruncles; fc area: frontal caruncle area; fc depth: frontal caruncle depth; strut rate: average pre-trial strut rate. Table II. Results after varimax transformation of a principal components analysis of the correlation matrix of five male characters as measured from 11 males Rotated factor pattern First factor Second factor Third factor Fourth factor Tarsus Spur length Stretched snood Skullcap width Caruncle area Proportion of variance explained by factor - 0.93 0.02 - 0.20 - 0.29 0.37 - 0.10 0.32 0.84 - 0.21 0.79 0.10 - 0.47 0.08 0.96 - 0.12 0.07 - 0.19 0.04 - 0.95 - 0.20 0.23 0.33 0.22 0.22 The rotated factor pattern indicates how strongly each character is associated with each independent axis. or greater forms of these characters, as hypoth- esized. Because females were used only once, male pairings were unique in every trial and the differ- ence between male characters was used in the analyses, each trial was independent despite the fact that most males were used more than once. I analysed the data using SASjmp (SAS 1989) and Statview (Abacus Concepts 1986) statistical soft- ware. I used one-tailed tests to analyse the females’ mating preferences, because both models of female choice clearly predict that females will prefer the more ornamented males. Where appropriate the results report mean values f SE. Results Eighteen of the 23 females tested signalled their choice of mates by soliciting copulation from one of the two males presented in each trial. Females strongly preferred males with comparatively higher scores on the second rotated component, that is, those with longer snoods and wider skull- caps (Table III; unpaired t-test: v= 17, x= 1.01, t=3.65, one-tailed P=O.OOl). The length of the male’s snood during display, the character state actually visible for assessment by females, was strongly correlated with the relaxed and stretched snood measurements (t-,=0.80, N=ll, P=O.Ol; r-,=0.75, N= 11, PzO.02, respectively). The other variables that loaded on the first, third and fourth axes (tarsus length, frontal caruncle area and spur length, respectively) did not account for female choice of males (t-tests: one-tailed DO.05). Char- acters analysed individually (pre-trial strut rate, trial strut rate, male condition, frontal caruncle depth, side caruncle number and beard length) Buchholz: Female choice in wild turkeys 93.5 Table III. Mean ( f SE) morphological and display measurements for males solicited by females and males not solicited Variable Solicited Not solicited P Tarsus 16.5 f 0.1 16.3 f0.1 NS Spur length 2.1 ho.0 2.2 f0.1 NS Beard length 13.6 f 1.5 14.5 f 16 Relaxed snood 3.3 rt 0.2 2.7 f 0.1 O.o”n”s Stretched snood 5.2 f 0.2 4.2 f 0.3 0.005 Side caruncles 31.0 * 3.4 24.0 f 36 NS Front caruncle area 20.7 f 1.6 17.7 LIZ 1.0 NS Front caruncle depth 1.3 zeo.0 1.2 f0.0 Skullcap width 4.9 f 0.2 4.2 zk 0.2 o.oNoss Mass 19.8 f 0.5 18.4 AO.5 Condition 1.2 50.0 1.1 *o.o 0.0”: First factor - 0.156 & 0.23 0.157 f 0.22 Second factor 0.524 f 0.2 1 - 0.48 1 f 0.20 o.Eo5 Third factor - 0.213 It 0.23 0.126 f 0.23 NS Fourth factor - 0.07 f 0.18 0.0 i 0.26 Displayed snood length 3.1 *0.2 2.2 + 0.2 0.:2 Pre-trial strut rate 3.4 i 0.5 3.0 f 0.6 NS Trial strut rate 0.2 ho.1 0,4 ho.1 Time with female 1.7 zrzo.7 0.4 f 0.2 O.o”o”l One-tailed probabilities are listed as not significant (NS) if P>O,O5. Wilcoxon signed-rank tests were used for direct measurements, and unpaired t-tests were used for the factor also did not explain female mating preference. (Wilcoxon signed-rank tests: P>O.O5). In the live male trials, females spent significantly more time with the male that they solicited (1.7 f 0.72 min for the chosen male versus 0.42 & 0.17 min for the male not chosen; Mann-Whitney U-test: U=74, U’=215, N,=17, N,=17, P=O.O2). The mean time until solicitation was 2.1 f 0.85 min. Time until solicitation was not dependent on the similarity of the scores or measures of any of the variables of the two males (Spearman rank correlation, P>O.O5). EXPERIMENT 2: MALE MODELS Methods Male models Each female was admitted singly to the mate choicearena, where she had a choice of interact- ing with either of two artificial males modified from plastic hunting decoys of hens. I altered these models to make them appear like strutting males.I painted the heads of the two decoys with enamel paint so that they resembled one another as closely as possible. I painted the skullcap light blue, the orbital area dark blue, the dewlap and throat red, and the sides and back of the neck pink with narrow red rings applied around the circumference. I removed the natural beards from two domestic turkey gobblers. I trimmed them so that they had an equal number of ‘hairs’, all 17 cm long, and attached them to the chests of the models. The decoys have a closed tail, typical of a resting bird. Strutting males, however, have a fanned tail. I constructed two artificial fans from coloured poster board, glue and black permanent marker. The vertical diameter of the tail was 33 cm and the horizontal radius was 69 cm. I positioned the fanned tails just behind the wing tips of the decoys and held them in place from behind with two thin wooden dowels that were stuck into the ground to stabilize the tail. The models themselves were mounted on a wooden dowel so that their backs were at a height of 065 m. Strutting turkeys drag the primary feath- ers of their wings on the ground, but I made no attempt to simulate this posture; the models had their wings in the normal resting position on their sides. 936 Animal Behaviour, 50, 4 To provide the sound of strutting, I played back a recording of a single strut from a com- mercially available video (GritTen Productions 1990). It was played back alternately between the males so that each male model ‘strutted’ four times per min. One 5-W speaker was placed to the side and behind each male and aimed towards the area immediately in front of the male. Using the individual volume controls on the speakers and a Tandy sound meter, I ensured that the amplitudes of the struts were equal (73 dB) at a point 0.75 m immediately in front of each male. The playback sounded slightly less loud to my ear than did the display of live males, but I did not measure the actual amplitude of male struts for comparison. The male models were identical in coloration, tail size, beard length and display frequency. To test whether the most variable types of fleshy head ornamentation seen in live males affect mate choice, I constructed artificial snoods and carun- cles for the artificial males. Four snoods (two 4.0 cm long, two 6.6 cm long), 14 side car-uncles (1.2 x 1.9 x I.0 cm) and four frontal caruncles (2 cm in diameter) were constructed from latex caulking. The snoods were painted as above so that they were bluish pink, and the cat-uncles of both types were painted red. These were applied with small pieces of double-sided tape so that one male model was ‘more ornamented’, with a long snood coupled with five caruncles on each side of his neck, whereas the ‘lessornamented’ male had the shorter snood and only two car-uncleson each side of his neck. Both males had two frontal caruncles placed side by side at the base of the front of their necks. I only varied snood length and side caruncle number becausethese characters were the most variable in wild-caught males. I exchanged one or more of the bodies, tails, beards, snoods or caruncles between the models after each trial, and randomized the position of the more ornamented male. Thus female prefer- ence could be attributed to the only consistent difference between the males: snood length and side caruncle number. Females I implanted 23 17-month-old female wild turkeys with a 21-day release pellet of 25 mg of oestradiol (Innovative Research of America, Toledo, Ohio) on 3 December 1992. After 10 days, when most females were exhibiting sexual behaviour, I conducted mate choice trials. Experimental design I conducted the male model trials between 0800 and 1600 hours from 13 to 15 December 1992. I started the playback of the strut recording before each female was admitted singly to the arena via a remotely opened door. Starting the moment the female stepped into the middle of the arena from the waiting area, I recorded the amount of time that she spent in each model’s half of the arena. A trial ended when the female solicited or after 20 min had passed. I tested each female only once. Data analysis If females choose mates at random with respect to male snood length and caruncle number, they should solicit an equal number of more orna- mented and less ornamented males. However, if females prefer males with the more elaborate form of these traits, they should solicit the more orna- mented male significantly more often. I used a binomial test (Siegel 1956) to test the statistical significance of the results. Rt?SUltS Of the 23 females tested in the male model trials, only nine solicited copulation from the decoys. One female immediately solicited in the centre of the entrance way of the arena; thus her choice was ambiguous and this trial was excluded. Of the eight remaining females, who clearly solicited on only one side of the arena, seven solicited before the more ornamented model (binomial test: one-tailed P=O.O4).Six of the eight females spent more time with the model male that they subsequently solicited; however, this differ- ence was not statistically significant (binomial test: one-tailed P=O+lS). The two females who did not spend more time with their preferred male also took much longer to choose (x=9 versus 1.7 min, respectively; overall f=4 min). Three of the females that chose the more ornamented model never spent any time on the side of the less ornamented decoy. The single female that solicited the less ornamented decoy never visited the more ornamented model, but solicited in a typical amount of time (1.9 min). Buchholz: Female choice in wild turkeys 937 Nine of the 14 non-soliciting females gave fre- quent alarm calls or frantically attempted to exit the arena, despite a training period. These females showedno preference for either side of the arena (binomial test: two-tailed P=O.S). The remaining fivefemaleswho neither solicited nor alarm-called engaged in non-sexual behaviour patterns such as dust-bathing or preening. They also showed no preferencefor either half of the arena (binomial test:two-tailed P= 1). CORRELATES OF ORNAMENTATION IN WILD MALES Methods I captured and measured free-living male wild turkeysto determine whether male ornamentation serves as a good indicator of male viability. Year- ling male wild turkeys were captured with bait druggedwith alpha chloralose (Williams & Austin 1988)at PaynesPrairie State Preserve in Alachua County, Florida between February and March of 1991 and 1992. I began trapping efforts several weeks before any males in the population began to displayand continued them until the approximate mid-point of the breeding season. 1 measured ornamentation as described above for captive males, with the exception of skullcap width, which I did not measure. I kept drugged birds indoors in boxes specially designed for transporting wild turkeys until they had recovered sufficiently for saferelease(2496 h). I counted the number of attached ticks on the head and neck and on the undersides of both wings. In addition I counted the lice on the feathersof the chest, back and rump by lifting and scanningthe feathers of each region for 2 min. I made blood smears from each individual using blood collected with a heparinized capillary tube from the alar vein. The smears were air-dried, tixedin 100% methanol, and stained. Blood para- siteburden was measured as the number of blood cellsinfected with haematozoa seen per 30 min of scanningunder oil immersion at 1000 x . Faecal samples were collected three times a day from the boxesof the recovering birds. I mixed 1 g from each sample with 20 ml of a saturated NaN03 solution and filtered it through cheesecloth into a centrifuge tube. The tube was filled to the very top so that a coverslip placed on the tube opening would remain adhered there during the subse- quent 5 min of centrifugation at 1500 rpm. During centrifugation, most parasite ‘eggs’ either float to the top and adhere to the coverslip or sink to the bottom, depending on their specific gravity. After centrifugation the coverslip was placed on a microscope slide and thoroughly scanned. I iden- tified parasite eggs and counted them directly or estimated very large numbers with a McMaster’s slide. At least six species of eimerian coccidia infect turkeys (Davidson & Wentworth 1992). Unfortunately, these speciesare difficult to iden- tify and could only be categorized as having large or small oocysts in this study. Large and small oocysts were similarly abundant in the samples, thus their numbers were summed for analysis. I thoroughly examined the sediment of each centri- fuge tube for trematode and other eggs under a dissecting microscope. I collected additional measures of ornamentation and blood smears from eight hunter-killed wild turkeys at Camp Blanding Wildlife Management Area, Florida, in March 1990. The potential of ornamentation to indicate male viability was assessedby testing for corre- lations between the measurements of each type of ornament and male age, condition and parasite burdens (within the yearling age class). Analysing these data proved problematic. Sample sizesvary because it was not possible to collect all the information from every male captured (e.g. faecal samples could not be collected from hunter-killed birds). Because of the low and varying sample sizes, a multivariate test using all male measure- ments was not tenable; instead, I calculated mul- tiple Spearman rank correlations. Interpretation of these correlations suffers from two problems. First, the low sample sizesmeant that the tests of significance were not very powerful (beta ranged from 0.45 to 0.72); thus it is likely that for some of the ornaments the null hypothesis may not have been rejected when it should have been. Power analysis revealed that sample sizesof 30 or more males would have been necessary to reduce the risk of type II error to acceptable levels (Brown et al. 1993). Second, conducting 30 statistical tests increases the collective risk of type I error to a value greater than 0.05. A very conservative way to correct this second problem is to divide the usual significance level by the number of tests conducted, resulting in an adjusted significance level of approximately 0.002. None of the corre- lations would be considered significant at this 938 , ~~wlal Behavior 50, 4 Table IV. Parasite loads of l-; zar-old, wild-c. 1:ht male wild turkeys .-- ’ I P>c;,n Occu rence ( SF) Parasite ,I II lad .~ Ticks Head Wing Total Lice Back Rump Total Eimeria oocyts Large Small Total Trichostrongylus eggs (total) Blood parasites Haemoproteus Leucocytozoon Plasmodium loci (9) 8.4 r 3.8 lot (9) 17.3 r 3.5 lOC> (9) 25.8 t 5.9 50 (8) 2.0 t 1.0 868(7) 6.8 t 2.3 lOtI (7) 7.0 :k 2.8 1001(9) 342.2 YIZ 190 89 (9) 2276.2 zk 2042 100 (9) 2618.4 i 2214 56 (9) 22.3 f 14.81 86 (7) 7.3 zt 2.5 14 (7) 0.1 * 0.1 43 (7) 0.6 zt 0.3 significance level, despite strong patterns of association in the data. I report the one-tailed probabilities of type I error in the results so that readers may evaluate the significance of the correlations themselves. Results The average parasite burdens of yearling male wild turkeys are presented in Table IV. Spearman correlations between male measurements and parasite burdens are reported in Table V. Sevenof the nine males sampled had attached ticks (Argasidae and Ixodidae). ‘Ticks were most com- mon on bare areas along the alar vein under the wing, but also occurred on the back of the head. The total tick burden was not significantly corre- lated with any measure of male ornamentation or body size. Lice were most common on the rump (only one of seven males had no rump lice). Four of the eight males sampled for back lice were infected. Total lice numbers were not correlated with any measure of male ornamentation, size or condition (all P>O.O5). One to five faecal samples were collected from each bird (x* s~z3.4 f 0.6). Faecal analysis revealed two common parasites: a protozoan coccidia, Eimeria spp., and a cecal nematode, Tr hostrongylus ten&. I also detected eggs from an inidentified cestode in one faecal sample flota- tior /. Sediments from the faecal samples appeared coi‘ lpletely free of parasite eggs. 1’111 nine males were infected with coccidia. Eiweria burdens were not negatively correlated wit I1tarsus length, mass or condition. The overall average coccidia burden showed a strong negative correlation with relaxed snood length (Fig. 3; ~a= - 0.73, N=9, P=O.O2). Individuals with extremely high burdens of coccidia in one sample (e.g. tens of thousands) never had samples with very low numbers of oocysts, and their lowest samples were usually higher than the highest burdens of individuals with lower average burdens (e.g. hundreds of oocysts). Relaxed snood length was not more strongly correlated with the lowest (or highest) coccidia load sampled from each bird than with the average total burden, although these measures tended to be negatively correlated with relaxed snood length (rs= - 0.55 to - 0.82, N=9, PcO.0