Birds were caught into mist-nets passively (HOSP) or using conspecific song playback recordings (other species). Sampling was carried out from Jan. to Oct. 2006 and each locality was sampled on a weekly basis. All captures took place between 05.00 and 11.30 h and no diel effects were observed for any variable examined. Birds were removed from the net within 3 min of capture and no more than 500 ml of blood was collected from the right jugular vein using a heparinized 0.3 cc syringe with a 29.5 gauge needle.

Approximately 5 ml of blood was used to prepare thin blood smears on glass microscope slides. Smears were air-dried at ambient temperature and stored until fixation. Males were identified by plumage in the case of HOSP and either by the presence of a developed cloacal protuberance in males during the breeding season or by unilateral laparotomy during the non-breeding season, for the other four species. Laparotomies were done under local lidocaine-induced topical anesthesia following guidelines approved by the Arizona State University Institutional Animal Care and Use Committee. During laparotomies we measured testis length to the nearest 1 mm. Age (hatchyear, HY; after-hatch year, AHY) was determined using flight feather and rectrix characteristics and the degree of skull pneumatization. We also recorded whether birds were molting. These measures enabled us to classify sampled birds as being in pre-breeding condition, breeding condition, or post-breeding condition (i.e., molting). Male birds were considered in breeding condition if their testis length exceeded half the maximum average length observed for each species during the breeding season. This threshold value was based on the conservative assumption that testes can produce sperm when at halfmaximal volume. To assess body condition, body mass (90.1 g) and tarsus length (91 mm) were measured. Birds were fitted with a uniquely numbered aluminum US Geological Survey leg band and were released at the capture site.

Blood parasites and leucocyte profiles

Blood smears were fixed for 10 min in absolute methanol within 5 days of collection and stained using the Giemsa method. Stained smears were dehydrated for one week under partial vacuum and then cleared using xylene. Slides were coverslipped and sealed using Cytoseal 60 (VWR, San Francisco, California) for long-term storage. Parasite prevalence, defined as the percentage of infected individuals in a sample, was determined by surveying blood smears at 250_magnification for 10 min and then at 400_magnification for 5 min using an Olympus BX60 light microscope (Olympus Optical Co., Tokyo). Slides were examined by a single observer (HBF) without knowledge of locality, species, or time of year that samples were collected. The density of infection with microfilariae and Trypanosoma was calculated as the number of parasites seen per unit of area. For this, we located large areas of the slide that had a relatively homogenous cell density and were about one cell layer thick. Then 200 randomly chosen microscope fields within this area were examined at 400_magnification and the number of each parasite type encountered was counted. To determine observer repeatability 12 randomly chosen slides from multiple species were each analyzed three times and parasite numbers were compared using Kendall�s coefficient of concordance. Only the number of microfilariae could be statistically tested and this was highly consistent (W_0.992, PB0.001). The number of parasite types detected on each smear was identical between observations.

To quantify the density of Haemoproteus infection, 25 non-overlapping microscope fields that contained nonoverlapping single cell layers were digitized at 400_ magnification. Image-Pro version 4.1 software (Media Cybernetics, Silver Springs, Maryland) was used to identify 10,000 erythrocytes per sample. Erythrocytes were identified based on morphological characteristics of their nucleus (aspect ratio, length, width, perimeter, roundness, relative brightness, and color) and were included in the counts if their entire nucleus was visible on the digitized image. Haemoproteus-infected erythrocytes seen on digitized images were manually counted and parasite densities were expressed as the number of Haemoproteus per 10,000 erythrocytes. The study of blood smears using light microscopy to detect blood parasites is not sensitive enough to conclude that individuals without visual evidence of parasites are parasite-free because subpatent infections may escape detection. Recently, Bentz et al. (2006) reported that blood smears may only detect relatively high parasitemia, suggestive of intense infection. Although this is an important concern, the above method does provides a relative measure of the degree of infection in a population which may ultimately have a bearing on the host biology, and has been widely used in studies of parasitism.

Differential leucocyte counts were determined by examining randomly selected and non-overlapping microscope fields of each smear at 1000_ magnification under oil immersion and counting leucocytes until we reached a total of 100 cells. Eosinophils (E), heterophils (H), basophils, lymphocytes (L), and monocytes were identified according to the criteria of Campbell (1995). The heterophil/ lymphocyte (H:L) ratio was used as an index of chronic stress (Vleck 2000, Bonier et al. 2007) based on the observation that increased glucocorticoid secretion may result in lymphocytopenia and a subsequent increase in heterophil numbers. Moreover, the H:L ratio increases in response to a variety of ��stressors�� including malnutrition, water deprivation, and injury. To determine repeatability of the leucocyte differential counts, 12 smears were analyzed three times each and Kendall�s coefficient of concordance was used to compare leucocyte numbers. Counts were highly correlated for the three most abundant leucocyte types (H: W_0.901, L: W_0.962, and E: W_0.862, all PB0.001). Total leucocyte count (TLC) was also measured to estimate the overall allocation to leucocyte production. For this the number of leucocytes per 10,000 erythrocytes were counted using the same digitized images generated for measuring Haemoproteus density.

Total corticosterone assay

Total CORT concentrations were quantified using commercial competitive enzyme-linked immunoassay kits (ELISA; Assay Designs Inc., Ann Arbor, Michigan, USA). This ELISA uses a polyclonal antibody with low cross-reactivity (less than 0.2%) with other steroids (manufacturer�s specifications). There was no difference for any species between the slopes of a curve produced by serial plasma dilution (2- to 32-fold) and a standard curve (all p > 0.3). Samples were assayed following a ten times dilution with assay buffer. All samples were assayed in duplicate and distributed randomly across assay plates. However, paired samples from a given individual (baseline and stress-induced) were assayed on the same plate. The optical density of assay wells was measured at 405 nm with a microplate absorbance plate reader. Plasma CORT concentrations were calculated via interpolation from the standard curve on the respective plate using GraphPad Prism vers. 4 (GraphPad Software Inc.: San Diego, California, USA). The sensitivity of the assay calculated from two standard deviations from a zero standard, ranged from 5.8�16.3 pg/ml and the mean intrassay coefficient of variation was 8.46% (n = 8 plates; 312 samples).

Corticosteroid binding globulin assay

Radioligand binding assays for CBG were based on Orchinik et al. (2000) and Deviche et al. (2001) with minor modifications. CORT binding capacity was determined using plasma stripped of endogenous steroids by incubation with 1% Norit A charcoal coated with 0.1% dextran in assay buffer (50 mM Tris acetic acid; pH 7.4) at 4 C or room temperature for 10 or 15 min, depending upon species. The incubate was then centrifuged for 10 min, and the supernatant was collected and then diluted with assay buffer (final dilutions vary between species). Radiolabelled CORT (1,2,6,7 3H-CORT, specific activity 70 Ci/mmol; Perkin Elmer Inc. Boston, Massachusetts, USA) was diluted in assay buffer and 50 ll of solution was dispensed into polypropylene tubes containing 50 ll of assay buffer (total binding) or of unlabeled competitor solution, and 50 ll of diluted steroid-free plasma. Nonspecific binding was determined in alternate samples using 3 lM of progesterone (P4), the hormone with the highest affinity for CBG (see results). Free and bound 3HCORT were separated by rapid vacuum filtration using a Brandel (M-24) harvester using glass fiber filters (GF-B) soaked for one hr in 0.3% polyethylenimine. Filters were immediately rinsed three times with 3 ml of ice-cold 25 nM Tris HCl buffer. The radioactivity in sample-soaked filters was measured by liquid scintillation, using a Beckman LS 6500 liquid scintillation b-counter (Beckman Coulter Inc., Fullerton, California, USA).

For each species, the equilibrium dissociation constant (Kd) and binding capacity (Bmax) of CBG were determined by incubating plasma pooled from five individuals with increasing concentrations of 3H-CORT (1 to 43 nM). Competition studies were performed using 1 nM 3H-CORT and the following unlabeled steroids: CORT, P4, dexamethasone (DEX), testosterone (T), 5adihydrotestosterone (DHT), and 17b-estradiol (E2) at concentrations ranging from 10 6 to 10 10 M. For these studies, samples were incubated at 4 C for 1 or 2 h, depending upon species. Estimates of potency (EC50) derived from competition experiments were converted to inhibition constants (Ki) using the equation of Cheng and Prusoff (1973). Binding parameter estimates were calculated using nonlinear regression models based on the law of mass action (GraphPad Prism vers. 4). Comparisons of best-fit nonlinear regression models were done using two-tailed Student�s t-tests with GraphPad Prism.

The CBG binding capacity of individual plasma samples was estimated using single point 3H-CORT binding assays, run in duplicate. Samples (10 ll plasma) were incubated with a total concentration of CORT equivalent to ten times the Kd to ensure that binding sites were nearly 90% saturated. To avoid using excessive amounts of radioactivity, this concentration was achieved using one part 3HCORT to five parts unlabelled CORT. The raw data were converted to Bmax by adjusting for percent saturation and the diluted specific activity of the radioligand. The Bmax values from single point assays and the species-specific Kd values derived from pooled samples were then used to estimate free CORT concentrations in plasma using the equation of Barsano and Baumann (1989).