We studied breeding P. h. mandellii birds at Shahetan station in the Lianhuashan National Nature Reserve (34°40′67″N, 103°30′84″E) during the April–July periods in 2009 and 2010. The elevation of the station is 2850 m asl. Mature mixed forests throughout the study area are dominated by spruce, fir, willow and birch trees (Sun et al. 2003). Local ground vegetation varies from sparse vegetation to dense shrubs of different species, as well as grasses, Arrow Bamboo (Sinarundinaria nitida), forbs and mosses (Sun et al. 2003).
We captured a total of 55 males using mist nets and marked them with unique combinations of colored leg bands to facilitate field identification. We captured male individuals as soon as possible when we detected their singing at the beginning of the breeding season (from late April to the middle of May). We opted for three types of measurements, i.e., body characteristics, ornamental and song characteristics to assess the relationship with female mate preferences. Specifically, we measured (1) bill length (exposed culmen), wing length, tail length and tarsus length as body characteristics using a centimeter ruler to the nearest 0.1 mm; (2) eyebrow size and wing-bar size (= length × width) as the ornamental characteristics using vernier calipers to the nearest 0.1 mm. We also recorded songs from differently marked males using a Sony WM-D6 recorder and a Sennheiser directional microphone. Song recording would be conducted at least 1 day after capture. A total of 139 recordings were analyzed using Avisoft SASLab Pro v. 4.52 applying the following settings: sampling frequency 22,050 Hz, 16 bit, time resolution 5.8 ms and bandwidth 162 Hz. In order to remove background sounds (the noises and sounds of other birds), all recordings were filtered using the FIR high-pass filter in the software with a low frequency limit of 4.5 kHz before our analyses. We measured the following song parameters: duration (DUR), interval (INT), maximum amplitude (MA), distance from start to maximum amplitude (DSM), maximum frequency in the maximum amplitude (MAXMA), minimum frequency in the maximum amplitude (MINMA), frequency bandwidth in the maximum amplitude (FBMA), maximum frequency (MAXF), minimum frequency (MINF) and frequency bandwidth (FB).
We found nests by following females during the nest building period (mainly from late May to early June) and checked daily to measure the length, width and mass of newly laid eggs. In order to minimize disturbance when following females, we would check the nest after the female left. This bird has strong territorial defense behavior throughout the breeding period. In order to determine the ownership of the nest by the male, we would identify the territory of each male bird according to our field observation and as well check the color rings of male birds to verify the ownership during the incubation and/or nestling period. All nests laid eggs successfully. We calculated the egg volume as 0.457 × length × width2/1000 (Hoyt 1979). Following Marchetti (1998), we used the first egg date as a quantitative measure of female mate preference (i.e., females preferred the characteristics of males who supported the earliest possible nestings). Since the nestling period of P. h. mandellii is about 14 days (Bi et al. 2009), we measured the nestlings on Day 12 (bill length, wing length, tail length and tarsus length) to assess the development during the nestling period.
In order to identify the relationships among the characteristics of the body of males, egg size and fledging size, we applied a Generalized Linear Model (GLM) using R (version 2.14.1). Since multi-collinearity of independent variables can cause problems in regression models (Hosmer and Stanley 2000), we first calculated the bivariate correlations among different variables and then used a correlation coefficient threshold of 0.7 as suggested by Fielding and Haworth (1995). Fifteen variables were retained for further analysis, i.e., bill length, wing length, tail length, weight, tarsus length, eyebrow size, wing bar size, DUR, INT, DSM, MA, MAXMA, FBMA, MINF and FB. We conducted simple regression models for these variables and excluded nonsignificant variables (p > 0.05). Subsequently, we used a multiple regression model containing all significant variables and applied a stepwise backward procedure based on Akaike’s information criterion (AIC, Akaike 1974) in order to reduce the number of variables.