Measurement parameters

We converted all recordings of great call climaxes into spectrograms (Window type: Hann, FFT size: 512 hz, frame overlap: 50 %), using Raven Pro 1.5 Sound Analysis Software (Cornell Lab of Ornithology Bioacoustics Research Program, Ithaca, New York), and made measurements from the spectrograms. The amplitude and frequency of the lar gibbon’s great call climax rises and falls over multiple notes [42, 49]; we defined it as consisting of three notes: the expiration note that reaches the highest frequency of the great call, and the notes that immediately precede and follow it (Fig. 1). We measured frequency and temporal parameters from 230 great calls (mean: 13.5 calls per animal; range: 2–27). The frequency measures were maximum and range (delta) of F0 frequency for the highest note of each climax, as well as the lowest F0 frequency of the great call climax. The temporal measure was the duty cycle of each climax, and was calculated as the total duration of the three notes in a climax (signal) divided by the signal plus the sum of the inter-note intervals that followed each climax note. A larger duty cycle thus represents longer climax note output relative to the breaks between notes. We also compared the total number of high notes (rapid inspirations) per call, across all animals.

Waveform (top) and spectrogram (bottom) of the lar great call, from a 24 year old female. Three high notes are indicated, and represent the sounds associated with inhalations between great call notes. The three notes of the climax are also labeled. Window type: Hann, FFT size: 1124 Hz, frame overlap: 50 %. Harmonics and background noise have been removed from this image

In addition, we combined the measures described above to create an index of call quality for each animal, to obtain a single measure that could be compared across all ages. We created the index by ranking animals, from 1 to 17 (n = 17 animals), in each of the first three measurement parameters (delta F0, maximum F0 and duty cycle). The lowest rank for a given measure, a rank of 1, was thus assigned to the animal that had the smallest value of that measure, and a rank of 17 (or the highest rank) to the animal with the highest value. Consistent with our predictions, high note production was ranked in the reverse order, so that the lowest ranked animal produced the greatest average number of high notes per call. All four ranking positions were then averaged for each animal, to produce its individual ranking index.

Two age groups were compared, using separate between-subject t-tests (two tails), to test the hypothesis that broadly differing age classes are distinguishable, even if more precise ages cannot always be predicted based on call characteristics. Too few subjects were available to divide the sample further. Thus, all animals below the median age of 17.6 years were categorized as young adults (n = 8, mean age: 12.9 years, range: 9–16.7 years), and all others as old adults (n = 9, mean age: 29.6 years, range: 17.6–45.2 years) (Table 1). For each of the measures we first calculated within-animal means, and then compared these means across all 17 animals. We corrected for multiple t-tests using the method described by Benjamini and Hochberg [56], which adjusted the P level of significance from 0.05 to 0.0375. In order to explore the overall relationships between age and the call parameters across the entire population, we also made separate regressions for each measurement parameter, as a function of caller age.

In order to determine how often great calls can be accurately assigned to either the young adult or old adult age groups, we also used the call parameters in a nested, permuted discriminant function analysis (pDFA) [57]. We conducted the pDFA using a function written in R [58] by R. Mundry, and based on the function Ida of the R package MASS [59]. Delta F0 was omitted as a classification parameter for this analysis, as it was highly correlated with the maximum F0 frequency measure. Three animals (D, E and G) were also excluded from the analysis because their call sample sizes were too small (2–3 calls). The remaining 14 animals had a mean of 15.9 ± 1.8 great calls per animal (range: 6–27).

In order to test the hypothesis that high notes correspond with an increase in vocal effort, we used separate paired T-tests to compare the delta F0, maximum F0, and the note duration of climax notes, depending on whether or not they were followed by a high note. We predicted that when a female exaggerates a climax note by increasing these parameters, it will require her to produce a rapid inspiration (high note), and thus high notes will follow longer notes that span a wider frequency range and reach a higher pitch.

The results of our comparison of song parameters by age allowed us to predict that these parameters would change over the course of song bouts, within individuals. We were able to record bouts containing ≥ 8 great calls from seven females, all from the wild population (mean: 9.9 calls per bout; range: 8–12, mean bout length 32.8 min, range 12–54 min). Two animals produced two bouts each; for each of these animals, we averaged the note parameters from each of their bouts and used those averages to represent that animal in subsequent comparisons. We calculated the mean delta F0, maximum F0, climax duty cycle and number of high notes per great call from each animal’s bout. Since bouts differed in the number of great calls produced, in order to generate regressions that compared measurement parameters over the course of bouts, we plotted each parameter over the last 7 great calls of a given bout, as a percentage of the average value obtained from the calls that preceded these 7 calls, at the start of the bout (n = 1–5 great calls, depending on the number of calls in a given bout). We also compared parameters obtained from the first four and last four great calls of bouts, predicting that delta F0, maximum F0, and duty cycle would be decreased, and high note production increased from early to late in bouts. We used matched-sample, two-tailed t-tests to compare the early versus late bout parameters.