Note Types and Coding in Parid Vocalizations: The ...

Note Types and Coding in Parid Vocalizations: The chick-a-dee Call of the Mexican Chickadee Poecile sclateri Author(s): Michele K. Moscicki, Marisa Hoeschele & Christopher B. Sturdy Source: Acta Ornithologica, 45(2):147-160. 2010. Published By: Museum and Institute of Zoology, Polish Academy of Sciences DOI: 10.3161/000164510X551282 URL: http://www.bioone.org/doi/full/10.3161/000164510X551282

BioOne (www.bioone.org) is an electronic aggregator of bioscience research content, and the online home to over 160 journals and books published by not-for-profit societies, associations, museums, institutions, and presses. Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of BioOne’s Terms of Use, available at www.bioone.org/page/terms_of_use. Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder.

BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research.

ACTA ORNITHOLOGICA Vol. 45 (2010) No. 2

Note types and coding in parid vocalizations: the chick-a-dee call of the Mexican Chickadee Poecile sclateri Michele K. MOSCICKI1, Marisa HOESCHELE1 & Christopher B. STURDY1,2,* 1Department of Psychology, P217 Biological Sciences Building, University of Alberta, Edmonton, Alberta, T6G 2E9, CANADA 2Centre for Neuroscience, P217 Biological Sciences Building, University of Alberta, Edmonton, Alberta, T6G 2E9, CANADA *Corresponding author: e-mail: [email protected]

Moscicki M. K, Hoeschele M., Sturdy C. B. 2010. Note types and coding in parid vocalizations: the chick-a-dee call of the Mexican Chickadee Poecile sclateri. Acta Ornithol. 45: 147–160. DOI 10.3161/000164510X551282 Abstract. To understand the communicative functions of any vocalization it is important to describe and classify the elements of that vocalization. Mexican Chickadees produce a namesake chick-a-dee call. Here, the note types (A, C, D, and Dh) from a sample of Mexican Chickadee chick-a-dee calls are identified, described, and classified. Frequency and temporal features of each note type are measured and compared to determine which features may be useful for note-type discrimination. Frequency measures, particularly peak frequency, appear to be most useful for discriminating among note types. Call syntax is analyzed to determine rules for note-type production. Mexican Chickadees produce the notes within their chick-a-dee calls in a consistent order: A → C → Dh → D with the potential for any note type to be repeated or omitted within this sequence. Similar to species in the brown-headed chickadee clade, B notes were not found in the calls of Mexican chickadees, suggesting this species may belong to the brown-headed clade. This work describes the chick-a-dee call of Mexican Chickadees and provides a foundation for future work aimed at understanding the communicative significance of this call within this species, as well as for comparative work on the chick-a-dee call among chickadee species. Key words: Mexican Chickadee, bioacoustics, Poecile, communication Received — June 2010, accepted — Oct. 2010

INTRODUCTION The use of oscine songbirds as a model for human language learning (Doupe & Kuhl 1999), and vocal communication in general (see Slater 2003 for a review), has become increasingly popular over the last few decades. One genus of songbirds, the North American chickadees Poecile, has received a great deal of attention in this regard. All members of this genus produce a namesake chicka-dee call. There is reason to believe that the chicka-dee call contains important species-relevant information, and as such it has been the subject of much research aimed at understanding the production, perception, and function of this vocalization (for reviews see Hailman & Ficken 1996, Lucas & Freeberg 2007, and Sturdy et al. 2007). Chick-a-dee calls are a common vocalization given by both males and females year round; they are composed of a number of discrete units (notes), thus allowing for many potential note

combinations and many potential encoded messages (e.g., Ficken et al. 1994). Syntactic evidence from Black-capped Chickadees Poecile atricapillus (Hailman et al. 1985, Hailman et al. 1987) and Mexican Chickadees Poecile sclateri (Ficken et al. 1994) reveals that the chick-a-dee call is an “open” call system such that one can find an increasing number of call types with increasing call sample sizes. Moreover, the chick-a-dee call is produced in a variety of contexts, such as mobbing (Templeton et al. 2005), contact (Smith 1991), flock identification (Nowicki 1989), and food location (Freeberg & Lucas 2002). Taken together, this evidence suggests that the chick-a-dee call is a vocalization that has the potential to convey a variety of important messages. Previous research on chickadee calls has revealed that different calls, and in fact different note types within calls, are used in different contexts. Black-capped Chickadees in close proximity to a predator produce calls with relatively more B

148

M. Moscicki et al.

notes, while calls with relatively more A notes are produced when farther from the same predator (Baker & Becker 2002). Black-capped Chickadee D notes may be important for flock identification (Nowicki 1989) and predator mobbing (Templeton et al. 2005) while Carolina Chickadee Poecile carolinensis D notes may play a role in recruiting flockmates to a food source (Mahurin & Freeberg 2009). In addition, Carolina Chickadees in flight produce calls with a high proportion of C notes (Freeberg 2008). Mexican Chickadees appear to use A notes during flight and movement, C notes in reaction to a disturbing stimulus, and D notes when perched (Ficken et al. 1994). It thus appears that closely related species use the note types in their chick-a-dee calls for a variety of different functions. Chick-a-dee note type production may play a functional role for many members of this genus and uncovering commonalities and differences in call and call note usage across Poecile species may lead to interesting insights about this vocalization. There are seven species of Poecile with varying degrees of relatedness. These species are generally divided into two sister clades: Black-capped, Mountain Poecile gambeli, and Carolina Chickadees are members of the black-headed clade, while Chestnut-backed Poecile rufescens, Boreal Poecile hudsonica, and Grey-headed Poecile cincta Chickadees are members of the brown-headed clade. The relationship of Mexican Chickadees to the other Poecile species is still debatable but this species is generally classified as a member of the brown-headed clade (Gill et al. 2005). Although a great deal is known about the genetic relatedness of the Poecile species, little is known about how these relationships affect the production and perception of Poecile vocalizations such as the chick-a-dee call. To date, detailed bioacoustic analyses have been performed on the chick-a-dee calls of four of the seven Poecile species (Blackcapped, Charrier et al. 2004, Mountain, Bloomfield et al. 2004, Carolina, Bloomfield et al. 2005, Chestnut-backed, Hoeschele et al. 2009). Previous work on the calls of Mexican Chickadees (Dixon & Martin 1979, Ficken 1990, Ficken et al. 1994) did not provide an in-depth analysis of the many individual note features present in these calls, such as frequency and duration measurements, which are necessary for more detailed comparative work. Although these species have been studied individually there has been no rigorous work comparing the chick-a-dee vocalization across species. This work is planned once the

chick-a-dee call of each of the seven Poecile species has been studied. The present studies focus on the chick-a-dee call of the Mexican Chickadee and have three main goals: i) to identify, describe, and classify the note types present in a sample of Mexican Chickadee calls in a manner analogous to other chickadee bioacoustic studies from our lab group; ii) to conduct a detailed analysis of frequency and duration measurements for each of the note types identified; and iii) to provide a syntactical analysis of Mexican Chickadee calls in a manner analogous to that provided for other chickadee species studied to date (i.e., Bloomfield et al. 2004, Charrier et al. 2004, Bloomfield et al. 2005, Hoeschele et al. 2009). This information will provide a thorough description of the Mexican Chickadee call and will serve as a foundation for future research designed to examine the perceptual and discrimination capabilities of this species. By using similar methodology to previous studies, this work can be used as a base for future in-depth comparative work on the chick-a-dee call of all the Poecile species.

MATERIALS AND METHODS We conducted three studies to fully describe the Mexican Chickadee call. In study one we classified Mexican Chickadee chick-a-dee call notes into note types; in study two we quantitatively measured notes from each note type to determine features useful for note-type classification; in study three we analyzed the syntax of the Mexican Chickadee chick-a-dee call. STUDY I — CALL NOTE CLASSIFICATION Recordings We obtained multiple recordings of Mexican Chickadee calls from three sources. We used recorded field notes to conservatively estimate that there was one bird vocalizing in each recording. Prior to any analysis, spectrograms of calls were visually assessed for recording quality; calls with excess noise, multiple birds calling simultaneously, or calls that were too faint to adequately classify note types were omitted from further analyses. Each recording contained a variable number of chick-a-dee calls. Following the removal of low quality call samples we were left with 235 calls from 10 recordings provided by the Macaulay Library of Natural Sounds at the Cornell Laboratory of Ornithology; 209 calls from

Mexican Chickadee chick-a-dee call

7 recordings provided by the Borror Laboratory of Bioacoustics at Ohio State University; and 99 calls from 5 field recordings made by Millicent Ficken. In total, we obtained 543 high quality chick-a-dee calls composed of a total of 1,360 individual notes. Recordings were made between July 1957 and July 1996 using either a Nagra III recorder and American D33 microphone; Marantz PMD-700 Recorder and Sennheiser ME-20 Microphone; Sony Walkman Professional cassette tape recorder and Electrovoice Soundspot microphone; or Sony Walkman Professional cassette tape recorder and Nakamichi CM-100 cardioid microphone. All recordings were from Portal, Arizona and were sampled at a rate of 44.1 kHz. We used a subsample of the recordings described above for analysis in Study I and Study II (below). The subsample was chosen by pseudorandomly selecting 10 recordings from the pool of 22 recordings described above. We pseudo-randomly sampled to ensure we included recordings from each of the three sources. Our subsample includes five recordings from the Macaulay Library, three recordings from the Borror Laboratory, and two recordings from Millicent Ficken. Ten calls were then randomly selected from each of these 10 recordings (100 calls, 320 notes) and included in the analyses for Studies I and II. Sound spectrograms and note classification Following procedures similar to those of previous bioacoustic studies (Bloomfield et al. 2004, Charrier et al. 2004, Bloomfield et al. 2005, Hoeschele et al. 2009), and to standardize analyses, individual notes were edited from the subsample of 100 calls and saved as sound files of 300 ms duration (non D-type notes) or 500 ms duration (D-type notes) by adding trailing silence to each sound file using SIGNAL 5.10.25 Sound Analysis Software (Engineering Design 2008, Berkeley, CA). Both D- and non D-type notes were saved as spectrogram images, with a cut-off amplitude of -35dB relative to the peak amplitude of the note, in a 512 point window (frequency precision = 86.1 Hz; temporal precision = 11.6 ms). This was done to achieve both adequate frequency and time resolution. Each of the note spectrograms were given a unique, random, four digit code for post-sorting identification and were printed, 15 spectrograms per page, on 8.5” x 11” glossy photo paper. Individual note spectrograms were cut out to form an individual note ‘card’ and thus produce a ‘deck’ of note spectrogram cards for sorting.

149

In a preliminary analysis, the first author sorted the deck of Mexican chick-a-dee note spectrogram cards into an unlimited number of categories based on visual similarity. Four note-type categories emerged during this process (see Fig. 1). A written description of each note type was prepared to provide two additional sorters with accurate information to use as a guide when independently sorting notes (see Note descriptions below). In addition, an exemplar of each category was provided; exemplars were selected from calls that had not been included in the current study. The descriptions, spectrogram card exemplars, and deck of note spectrogram cards were given to the second and third authors to sort; all three sorters had previous experience sorting call notes of other chickadee species. Once all three sorters had sorted the entire deck of note cards into note-type categories, the percent agreement among sorters was calculated to determine the reliability of the note-type categories. A meeting was held to resolve any disagreements in note sorting and refine note sorting criteria. STUDY II — NOTE MEASUREMENTS The same subsample of 320 Mexican Chickadee notes from Study I was used in the present study. We measured a total of 85 A notes, 134 C notes, 56 D notes, and 45 Dh notes. SIGNAL 5.10.25 Sound Analysis Software was used to make all note measurements following similar procedures used to analyze the chick-a-dee call notes of other species (e.g., Bloomfield et al. 2004, Charrier et al. 2004, Bloomfield et al. 2005, Hoeschele et al. 2009). Frequency measurements were taken on spectrograms with a Hanning window size of 1024 points with a frequency precision of 43.1 Hz. Temporal measurements were taken on spectrograms with a Hanning window size of 256 points with a temporal precision of 5.8 ms. We measured different frequency and duration parameters for each note type depending on the acoustic characteristics of the particular note type (i.e., tonal or broadband). For A, C, and the A note-type portion of Dh notes, we analyzed the following frequency measurements on the frequency band with the highest amplitude (i.e., the darkest band in the spectrogram): start frequency (SF), peak frequency (the highest frequency of the loudest frequency band; PF), and end frequency (EF; see Fig. 2A). Temporal measurements on these note types consisted of ascending duration (AD) and descending duration (DD). We used these

150

M. Moscicki et al.

Fig. 1. A sound spectrogram (FFT window = 512 points, frequency precision = 86.1 Hz, temporal precision = 11.6 ms) showing exemplars of each of the four note types identified in our sample of Mexican Chickadee chick-a-dee calls.

measurements to calculate the slope of the ascending (FMasc) and descending (FMdesc) frequency modulations for these note types. Frequency modulation is calculated as follows: FMasc = (PF-SF)/AD and FMdesc = (PF - EF)/DD. Because A notes and the A note-type portion of Dh notes have many oscillations of ascending and descending frequency modulation, we measured AD and DD on the oscillation with the greatest frequency range. We calculated total duration (TD) for all note types (A, C, D, and Dh; see Fig. 2B). For all note types, we generated a spectrum (window size of 4,096 for C notes and 16,384 for A, D, and Dh notes) that encompassed the entire note and had a smoothing width of 88.2 Hz. This spectrum was used to calculate the frequency at maximum amplitude (i.e., the loudest frequency; Fmax; see Fig. 2C) for all note types. We also used

this spectrum to calculate the fundamental frequency (i.e., the lowest frequency within 35dB of the peak amplitude of the note; f0) and the note peak frequency (i.e., the highest frequency within 35dB of the frequency with peak amplitude; NPF) for C notes, D notes, and the D note-type portion of Dh notes (see Fig. 2D). Statistical analyses In order to determine whether certain features of a note can be used to classify that note into a certain note-type category, we examined the variation of the measured note features within each note type and across all note types. Following the methodology used to analyze the features of vocalizations in several other Poecile species (e.g., Bloomfield et al. 2004, Charrier et al. 2004, Bloomfield et al. 2005, Hoeschele et al. 2009), we

Mexican Chickadee chick-a-dee call

151

Fig. 2. A sound spectrogram (FFT window = 512 points, frequency precision = 86.1 Hz, temporal precision = 11.6 ms) and spectra illustrating measurements of note-type features. Panels A, B, and C feature a representative A note as an example while panel D features a representative C note. A) Spectrogram showing frequency measurements used on A, C, and the tonal portion of Dh notes. SF — start frequency; PF — peak frequency; EF — end frequency. B) Spectrogram showing duration measurements; vertical lines indicate the boundaries of duration measurements. AD — ascending duration; DD — descending duration; TD — total duration. C) Spectrum (window size = 16,384 points) showing the frequency at maximum amplitude (Fmax indicated by the vertical line). D) Spectrum (window size = 4,096 points) indicating the lowest frequency above -35dB relative to the peak (f0; indicated by the vertical line; -35dB relative to peak indicated by the solid horizontal line), Fmax and the highest frequency above -35dB from peak (NPF; indicated by the vertical line).

calculated a measure of potential for note-type coding (PNTC). PNTC provides a quantitative measure of how different each feature is between note types and can be used to compare the magnitude of feature differences across note types as well as across species. PNTC is derived from a method used to assess the potential for individual coding (PIC). PIC has been widely used in the literature to determine what features of a vocalization may be used to identify individuals (see Lengagne et al. 1998). We have adapted this method to search for features that may be useful for identifying note types rather than individuals.

PNTC is calculated separately for each note feature by dividing the coefficient of variation for that note feature between note types (CVb) by the mean coefficient of variation for that note feature within note types (mean CVw). The idea behind this measure is that if the variation of a note feature is greater across note types than within a note type then the PNTC value will be > 1 and that feature may be useful for note-type discrimination. Alternatively, if variation on a feature is greater within note types than across note types then PNTC will be < 1 and that feature is likely less useful for note-type discrimination.

152

M. Moscicki et al.

The coefficient of variation between note types is calculated as follows: CVb = (SD/χ) × 100, where SD is the standard deviation and χ is the mean of the note feature measure calculated across all note types (i.e., one standard deviation and mean are calculated using all note types, A, C, D, and Dh). The mean coefficient of variation within note types is calculated using the same formula, CVw = (SD/χ) × 100; however, in this calculation SD is the standard deviation and χ is the average of the note feature measure within one note type (i.e., one standard deviation and mean is calculated for each note type separately, A, C, D, and Dh). CVw is then calculated for each note type separately (resulting in four CVw calculations). These CVw calculations are then averaged to obtain the mean CVw. PNTC equals the ratio CVb/mean CVw. As a compliment to our PNTC analyses, we also conducted a linear discriminant analysis (LDA) using Wilks’ stepwise method (SPSS v. 15.0.0; SPSS Inc. 2006) to determine which note features have the potential to be important for correctly classifying Mexican Chickadee note types. We set the criteria for entry and removal from the model at p < 0.05 and p > 0.10, respectively. In addition to the LDA, we performed a classification tree analysis using the rpart package for the statistical program R (Therneau & Atkinson 2010). This non-parametric test circumvents the assumptions inherent in the parametric LDA. Both the LDA and classification tree models were cross-validated. We also used SPSS version 15.0.0 to perform univariate analyses of variance (ANOVAs) in order to assess differences among note types on the note features we measured. In total, we performed 11 separate ANOVAs, one for each note feature measured. Because we conducted multiple comparisons, we used Bonferroni corrections to control type I error. We measured 9 note features for A notes, 11 note features for C and Dh notes, and 5 note features for D notes; thus, our α levels were set at α = 0.05/9 = 0.0056 for A notes, α = 0.05/11 = 0.0045 for C and Dh notes, and α = 0.05/5 = 0.01 for D notes. We used the GamesHowell post-hoc test for samples where equal variance is not assumed. Bonferroni corrections were used for all post-hoc tests depending on the number of comparisons used. Because some note features were not measured for all note types, the Bonferroni corrections were either α = 0.05/3 = 0.017 (for SF, PF, EF, DD, FMasc, FMdesc, f0, and NPF) or α = 0.05/4 = 0.0125 (for AD, TD, and Fmax).

STUDY III — SYNTACTICAL ANALYSIS Recordings and sound spectrograms The entire sample of Mexican Chickadee chicka-dee calls described in Study I was considered for analysis in this study and consisted of 543 calls composed of 1,360 notes. Each whole call was saved to a separate sound file using Syrinx v.2.6h software (Burt 2006). Spectrograms of individual calls were then made using SIGNAL 5.10.25 Sound Analysis Software. Spectrograms were 1,800 ms in duration in a 512 point Hanning window. The cut-off amplitude for all spectrograms was set to -40dB relative to the peak amplitude of the call and each individual whole call spectrogram was saved as an image file. Call spectrograms were printed, four spectrograms per page, on 8.5” ×11” white paper and given a unique file name for later identification. Call syntax determination The same sorters from Study I determined call syntax in the whole call spectrograms by sorting notes into note-type categories using the category descriptions and exemplars described in Study I. All note sorters were blind to the classifications of the other sorters. The note classification procedure was slightly different between this study and Study I (seeing notes in the context of the whole call in the former and seeing notes in an isolated context in the latter). We compared the classification of notes used in Study I to the classification of those same notes using the methodology of this study (i.e., notes sorted visually but seen within the context of the whole call); this was done to ensure that the different note classification methods would not lead to different results. Percent agreement among the three sorters was determined and a meeting was held to discuss any discrepancies in note classifications. Probability calculations Once note types were determined for all 1,360 notes in the sample, the number of different call syntax types produced in this sample, and the number of times they occurred, was determined. To examine the syntactical rules of the Mexican Chickadee chick-a-dee call, conditional transition probabilities for all syntax types were calculated. This measure involved calculating the probability that a certain note type would occur given that a certain note type had directly preceded it; for example, the probability was calculated that a D note would occur given that an A note was

Mexican Chickadee chick-a-dee call

produced directly before it in the call. In order to calculate this measure, the number of transitions from the note type of interest to the second note type of interest was counted (e.g., the number of transitions from an A note to a D note) and this was divided by the total number of transitions from the preceding note type (A notes in this example). This was done for all note-type combinations as well as the probability that the note was the last note in the call (i.e., that it was followed by no note type). We also calculated two note position probabilities. First, we calculated the probability that a certain note type occurred in a certain position within the call (e.g., the probability of an A note occurring in the 1st position in the call). To calculate this probability, the total number of the note type of interest (in this example, A notes) in the position of interest (in this example, the 1st position in the call) was summed and divided by the total number of times that note position occurred in the call sample. The 1st note position would occur in all calls; however, if a call was only 3 notes long then the 4th note position onwards would not occur in that call. The second note position probability we calculated was the probability that a certain position in the call contained a certain note type (e.g., the probability that the 1st position in a call contained an A note). To calculate this probability, instead of dividing the number of A notes that were in the 1st position in the call by the number of 1st positions in our call sample (as in the first probability we calculated), the number of A notes in the 1st position in the call were divided by the total number of A notes in all calls in our sample.

RESULTS STUDY I — CALL-NOTE CLASSIFICATION Four note-type categories (see Fig. 1) were identified by sorting the 320 Mexican Chickadee note cards. We based the nomenclature for our four note-type categories on note names previously described by Ficken (1990), as well as on similar note types appearing in calls of related chickadee species that have been well studied (e.g., Blackcapped Chickadees, Charrier et al. 2004). The percent agreement among the three sorters was 98.73% (four note disagreements) before a meeting was held to revise the notetype category criteria. The criteria for Dh notes were refined to ensure that Dh notes with a short

153

terminal D note-type portion were not misclassified as A notes. This clarification resolved the four disagreements and percent agreement among sorters reached 100%. Note-type descriptions A Notes: A notes are tonal (i.e., no overtones or harmonic-like bands), as with the A notes of other chickadee species, and generally occur between 6–8 kHz. These notes have multiple discrete cycles of ascending and descending frequency modulation throughout their duration; this is different from the A notes of all other chickadee species studied to date. There is a very slight decrease in frequency from note start (about 8 kHz) to note end (about 6 kHz; see Fig. 1). These notes are typically between 150–300 ms in duration. C Notes: C notes cover a wider frequency range than A notes (typically covering about 6 kHz with frequencies between roughly 2–8 kHz). These notes have ‘stacks’ of relatively parallel frequency bands. The main frequency band begins at a low frequency (around 4 kHz) and increases generally linearly, to a peak frequency (around 5 kHz) and either ends at this peak frequency or decreases very slightly in frequency before note termination (see Fig. 1). All other frequency bands follow this general pattern of relatively linear frequency increase to a peak where they either terminate or decrease in frequency slightly before termination. C notes typically range from 40–60 ms in duration. D Notes: D notes appear as broadband noise with an underlying banded structure. These notes are longer in duration than both A and C notes and range from 350–500 ms. D notes typically cover a frequency range of about 5 kHz with frequencies ranging from roughly 3–8 kHz, although this can be quite variable. There is little to no frequency modulation throughout the D note (see Fig. 1). D hybrid (Dh) Notes: Dh notes appear as a combination of an A note-type portion followed by a D note-type portion. The A note-type portion has discrete cycles of ascending and descending frequency modulation (between about 6–8 kHz) similar to an A note and typically spans a duration of between 100–150 ms. This portion terminates with a rapid decrease in the frequency of the tonal band over a short duration (see Fig. 1). The D note-type portion of the Dh note is continuous with the initial tonal frequency band and appears as broadband noise between about 3–6 kHz that decreases in bandwidth over roughly 50 ms and terminates with a narrower frequency

NPF f0

4523.3 ± 121.5 2764.1 ± 65.7 23.6 32.3 15.3 26.6 1.5 1.2 112.2* 21.0* 3, 316 2, 232 ** ** except between A and C p = 0.03

Fmax FMdesc

70.8 ± 4.9 103.4 56.4 1.8 545.7* 2, 261 ** 74.5 ± 5.3 95.5 62.7 1.5 35.4* 2, 261 **

FMasc TD

161.9 ± 7.5 82.6 26.1 3.2 416.9* 3, 316 ** except between D and Dh p = 0.10 10.3 ± 0.4 59 38.6 1.5 220.4* 2, 261 ** except between A and Dh p = 0.35

DD AD

21.9 ± 1.1 89 31.1 2.9 262.2* 3, 316 ** except between A and Dh p = 0.96 4184.1 ± 92.9 35.8 13.6 2.6 1145.9* 2, 261 ** except between C and Dh p = 0.09

EF PF SF

Table 1 shows a summary of the PNTC and ANOVA results for all note features measured on each note type. The PNTC results show that all note features measured, except for note peak frequency (NPF), have PNTC values > 1 and thus could potentially be used to discriminate among note types. Peak frequency (PF) has the highest PNTC value, at 6.26, and seems the most likely feature to use for note-type discrimination. Descending duration (DD), both frequency modulation measures (FMasc and FMdesc), as well as all power spectra measurements (Fmax, f0, and NPF) appear to be less useful for determining note type as their PNTC values are much closer to 1 than any of the frequency measures (SF, PF, EF), ascending duration (AD), or total duration (TD). All of the latter note features have PNTC values greater than 2. The LDA results indicate that Mexican Chickadee chick-a-dee note types are distinct. In total, 99.4% of the sample was correctly classified by the LDA; only two of 320 notes were misclassified. One A note was misclassified as a C note and one Dh note was misclassified as an A note. The cross-validation results are identical to the original LDA results. The stepwise model stopped after inclusion of eight of the 11 note features (FMasc, DD, and Fmax were not included in the model). PF, EF, and TD were the first three note features used in the stepwise model and together indicate that group means (i.e., note types) are not equal (χ2(24, N=320) = 2712.745, p < 0.001). We also examined the standardized discriminant function coefficients and structure coefficients of the first discriminant function in our LDA. Peak frequency (PF) had both the highest standardized discriminant function coefficient (0.771) as well as the highest structure coefficient (0.920) in the first discriminant function. This indicates that peak frequency is the most important contributor to the first discriminant function (Betz 1987). This agrees with our PNTC analysis which indicated that peak frequency may be the most important feature for distinguishing among Mexican Chickadee note types. Our classification tree shows that three decision rules are needed to classify all 320 notes in our sample with 100% accuracy (see Fig. 3). The

4908.9 ± 175.1 5916.7 ± 157.8 57.9 43.3 11.5 6.9 5 6.3 1630.1* 4180.2* 2, 261 2, 261 ** ** except between A and Dh p = 0.03

STUDY TWO — NOTE-TYPE ACOUSTIC FEATURES

Mean ± SEM CVb Mean CVw PNTC F df Games-Howell Test

band (between about 3–5 kHz) for a D note-type portion duration between 100–200 ms. The entire note spans a duration of 300–500 ms.

15286.3 ± 368.0 36.9 39.3 0.9 32.7* 2, 232 ** except between A and Dh p = 0.54

M. Moscicki et al. Table 1. Potential for note-type coding (PNTC) for the 11 features measured on Mexican Chickadee note types (A, C, Dh, and D). SF — start frequency; PF — peak frequency; EF — end frequency; AD — ascending duration; DD — descending duration; TD — total duration; FMasc — ascending frequency modulation; FMdesc — descending frequency modulation; Fmax — frequency at maximum amplitude; f0 — fundamental frequency; NPF — note peak frequency. * — Signifies notes were significantly different. ** Signifies significant differences among all note pairs in Games-Howell post-hoc tests. Exceptions between pairs that were not significant in the Games-Howell post-hoc analyses for each note feature are indicated below the asterisks and are followed by their corresponding non-significant p-values.

154

Mexican Chickadee chick-a-dee call

155

STUDY III — SYNTACTICAL ANALYSIS

TD < 72.9 ms

f0 < 469.8 Hz C

SF < 3114 Hz A

D

Dh

Fig. 3. A classification tree showing that all 320 Mexican Chickadee notes can be separated accurately into their respective categories using only three decision rules. The first rule indicates that all note types with a total duration (TD) less than 72.9 ms in duration are C notes. Of the remaining A, D, and Dh notes, all notes with a fundamental frequency (f0) less than 469.8 Hz are A notes. The last decision rule shows that D notes can be separated from Dh notes using start frequency (SF): D notes have a start frequency less than 3114 Hz whereas Dh notes have a start frequency greater than 3114 Hz.

first rule separates notes that are less than 72.9 ms in total duration (TD) from the rest of the notes. This rule separates all C notes from the rest of the note types. Rule two separates notes that have a fundamental frequency (f0) less than 469.8 Hz from all other note types. This rule separates all A notes from the remaining note types. Finally, rule three separates D notes from Dh notes with the former having a start frequency (SF) less than 3114 Hz and the latter having a SF greater than 3114 Hz. Our ANOVA results indicate that all note types differed significantly on every note feature measured (Fs ≥ 21.023, all ps ≤ 0.001; see Table 1). The results of the Games-Howell post-hoc test show that a few note types were not significantly different on a subset of the note features measured. Bear in mind that all post-hoc α levels have been Bonferroni corrected to either 0.017 or 0.0125 based on the number of note types that were measured for a specific feature (see Statistical analyses above). See Table 1 for a list of which features did not differ significantly in post-hoc tests. All other notes differed on all features measured (mean differences ≥ 8.3, ps ≤ 0.015; see Table 1 for F values).

Note-type percent agreement among sorters for all 1,360 notes was initially 92.06% (108 note disagreements). There were no disagreements between the classifications of notes from Study I and the classification of those same notes in Study III. This suggests that seeing the note in the context of the whole call did not affect the ability of experienced sorters to classify the note type. After meeting to resolve note-type sorting discrepancies we further refined the description of Dh notes to ensure that A and D notes in close temporal proximity were not misclassified as Dh notes (see Note descriptions above). This refinement brought the percent agreement among all note sorters to 98.68% (18 note disagreements). The first author made the final note-type decision on these last 18 notes. The number of call types in a sample has been defined in the literature in one of two ways: call types are either reported using an expanded method with repetitions of the same note type indicating a different call type (i.e., ACCD is a different call type than AACCCDD), or using a condensed method and disregarding repetitions of same note types and lumping repetitious calls into one call type (i.e., ACCD and AACCCDD are the same call type, namely, A, C, D). Previous work by Ficken (1990) examining Mexican Chickadee calls and work by Hoeschele et al. (2009) on the Chestnut-backed Chickadee of the brown-headed clade use the first convention for call type reporting. Charrier et al. (2004) and Bloomfield et al. (2004, 2005) use the second method of lumping call types when analyzing the syntax of birds of the black-headed chickadee subgroup. We have thus reported our findings using both methods so that our work may be comparable to all previous works mentioned above. In total, we identified 33 different expanded syntax types and 11 condensed syntax types in our sample of Mexican chick-a-dee calls (see Tables 2A and B for a detailed list of call types reported in each manner described above). All calls in this sample, except one, followed a pattern of note production; notes were always produced in the order A → C → Dh → D within a call. Some note types were omitted or repeated within certain calls, but this rule remained true for 542 of 543 calls. The one remaining call used syntax DhACCC. Table 3 shows note-type transitions along with conditional transition probabilities. The most

M. Moscicki et al.

156

Table 2. Syntax types observed in our sample of Mexican Chickadee chick-a-dee calls. A — Calls with repeated note-types are considered different syntax types. B — Calls with repeated note types are considered the same syntax type.

A Syntax A AA AAA AAAAA AAAD AAADD AACCDh AAD AADD AC ACC AD ADD ADh ADhD ADhDD ADhDDD C CC CCC CCCC CCCCC CCCCCC CCCCCCC CCCCCCCC CCCCCDD Dh DhACCC D hD DhDD DhDDD DhDDDD D hD h

Number of Calls

% of Sample

17 61 10 1 9 2 1 21 3 2 32 89 20 40 14 1 1 1 18 12 10 5 11 7 6 1 45 1 77 19 3 1 2

3.13% 11.23% 1.84% 0.18% 1.66% 0.37% 0.18% 3.87% 0.55% 0.37% 5.89% 16.39% 3.68% 7.37% 2.58% 0.18% 0.18% 0.18% 3.31% 2.21% 1.84% 0.92% 2.03% 1.29% 1.10% 0.18% 8.29% 0.18% 14.18% 3.50% 0.55% 0.18% 0.37%

Number of Calls

% of Sample

89 144 1 34 40 16 70 1 47 1 100

16.39% 26.52% 0.18% 6.26% 7.37% 2.95% 12.89% 0.18% 8.66% 0.18% 18.42%

B Syntax A A, D A, C, Dh A, C A, Dh A, Dh, D C C, D Dh Dh, A, C D h, D

common transitions we observed were C notes followed by other C notes (19.72% of all transitions), D notes as the final note in a call (19.35% of all transitions), and A notes followed by D notes (10.60% of all transitions). These results are highly similar to results reported by Ficken et al. (1994)

regarding the syntax of the Mexican Chickadee call. Ficken et al. (1994) found that A notes most often transit to D notes, D notes repeat or end a call, and C notes most commonly follow C notes. When we examine conditional transition probabilities, we see that A notes can be followed by all other note types, as well as be the final note in a call. C notes are more rigid in that 99.46% of the time they are either followed by another C note or are the final note in the call. This pattern holds true for Dh and D notes as well, with 98.55% of Dh notes being followed by another Dh note or being the last note in a call, and 100% of D notes being followed by another D note or being the last note in the call. Compared to other chickadee species, the chick-a-dee call of the Mexican Chickadee is relatively short (Bloomfield et al. 2004, 2005, Charrier et al. 2004). Only 34.81% of calls in this sample were 3 notes or longer. When examining the note position probabilities, we discovered that when A notes are present they are usually the first note in the call (70.39%); this is also true for Dh notes (71.50%). When C notes are present they are most often the second note in a call (27.12%); this is also true of D notes (63.75%; see Table 4). This distribution of note types further reflects the brevity of Mexican Chickadee chick-a-dee calls.

DISCUSSION In these studies, we have conducted an indepth analysis of the chick-a-dee call of a sample of Mexican Chickadees. We have identified, described, and classified the note types, measured individual note features, and analyzed overall syntax of this call. We show that Mexican Chickadees produce at least four distinct note types (A, C, D, and Dh notes). Our analysis of individual note features indicates that there are many features that may be useful for discriminating among note types. Finally, we show that the notes in this species’ chick-a-dee call are arranged in a consistent syntactical order of the pattern A → C → Dh → D. These findings are in line with those reported for other chickadee species studied to date (e.g., Bloomfield et al. 2004, 2005, Charrier et al. 2004, Hoeschele et al. 2009) and will further our understanding of the complex chick-a-dee vocalization of the Poecile species. In our sample of Mexican Chickadee calls, we were able to define note categories and classify notes into these categories with high agreement

Mexican Chickadee chick-a-dee call Table 3. Transitions from one note type to another in our sample of chick-a-dee calls from Mexican Chickadees. A — denotes the total number of transitions from each note type to each other note type. B — denotes the conditional transition probabilities (i.e., the probability that one note type may occur given that a certain note type has occurred just previously).

A Transitions

Number

% of Sample

132 36 144 56 91 0 268 1 1 103 0 0 57 0 263 1 0 116 2 88

9.71% 2.65% 10.60% 4.12% 6.70% 0.00% 19.72% 0.07% 0.07% 7.58% 0.00% 0.00% 4.19% 0.00% 19.35% 0.07% 0.00% 8.54% 0.15% 6.48%

A->A A->C A->D A->Dh A-> no note C->A C->C C->D C->Dh C-> no note D->A D->C D->D D->Dh D-> no note Dh->A Dh->C Dh->D Dh->Dh Dh-> no note

B Conditional Transitions Given A first Total of 459 transitions from A

A->A A->C A->D A->Dh A-> no note Given C first C->A Total of 373 transitions from C C->C C->D C->Dh C-> no note Given D first D->A Total of 320 transitions from D D->C D->D D->Dh D-> no note Given Dh first Dh->A Total of 207 transitions from Dh Dh->C Dh->D Dh->Dh Dh-> no note

Conditional Percentage 28.76% 7.84% 31.37% 12.20% 19.83% 0.00% 71.85% 0.27% 0.27% 27.61% 0.00% 0.00% 17.81% 0.00% 82.19% 0.48% 0.00% 56.04% 0.97% 42.51%

among note sorters. However, in order for these categories to be ecologically valid they must be verified by the Mexican Chickadees. This can be accomplished using operant experiments such as those previously used successfully with other songbird species (e.g., Sturdy et al. 1999, 2000).

157

Once we have established that our note-type categories are valid in the perceptual worlds of the Mexican Chickadees, we can then explore more fine-scale mechanisms of note-type perception. Our PNTC analyses indicate that start frequency may be an important feature for distinguishing between Mexican Chickadee note types. A similar prediction stemming from PNTC analyses of Black-capped Chickadee introductory notes was tested using operant experiments and the results appear to support the PNTC-driven predictions that start frequency is important for note type discrimination in Black-capped Chickadees (Charrier et al. 2005). The usefulness of predictions from PNTC analyses in Black-capped Chickadees leads us to believe that our PNTC predictions will also provide insight into the perceptual mechanisms of Mexican Chickadees, or at least provide a starting point for perceptual research in this species. Our LDA and classification tree analyses support our PNTC predictions that note features such as start frequency, peak frequency, and total duration may be important for note-type discrimination in Mexican Chickadees. Our PNTC, LDA, and classification tree analyses all indicate that Mexican Chickadee notes are very distinct from one another. This appears to be in contrast to the introductory note types of many of the Blackheaded Chickadees species where note types are so closely related that there are even hybrid note types because the distinction between categories of notes is unclear (i.e, A/B notes in Mountain Chickadees; Bloomfield et al. 2004). Determining the cause of these note type differences among related species is an area ripe for future investigation. The Mexican Chickadee note types we observed were similar, but not identical, to those reported for the chick-a-dee calls of other species. For instance, Mexican Chickadees produced an A note that was acoustically distinct from A notes produced by other chickadee species. In fact, the Mexican Chickadee A note is much more similar to the Buzz note produced by the Brown Creeper Certhia americana and the Zeet note produced by the Golden-crowned Kinglet Regulus satrapa than to the A note produced by any other chickadee species (Ficken 2000). The Mexican Chickadees in our sample flock closely with both Brown Creepers and Golden-crowned Kinglets. It will be interesting to determine if Mexican Chickadees that do not flock closely with these heterospecifics produce A notes less similar to the Buzz of Brown Creepers and the Zeet of Golden-crowned Kinglets

M. Moscicki et al.

158

Table 4. Positional note probabilities: A — the probability of each note type in each position in the call in our sample of Mexican Chickadee chick-a-dee calls (i.e., the probability that the 1st note in a call will be an A note). Total number of calls in parenthesis. B — the probability of the position of each note type within our Mexican Chickadee chick-a-dee call sample (i.e., the probability that an A note will be the 1st note in a call) as well as the number of calls in our sample containing each note type.

A

Probability A Note C Note D Note Dh Note Probability Note Position Occurs

Note 1 (540)

Note 2 (477)

Note 3 (188)

Note 4 (63)

Note 5 (38)

Note 6 (30)

Note 7 (17)

Note 8 (6)

59.44% 12.78% 0.37% 27.41%

22.85% 20.75% 44.23% 12.16%

11.70% 44.15% 44.15% 0.00%

1.59% 66.67% 31.75% 0.00%

7.89% 76.32% 13.16% 2.63%

0.00% 70.00% 30.00% 0.00%

0.00% 94.12% 5.88% 0.00%

0.00% 100.00% 0.00% 0.00%

100.00%

88.33%

34.81%

11.67%

7.04%

5.56%

3.15%

1.11%

B

A C D Dh

Note 1

Note 2

Note 3

70.39% 18.90% 0.60% 71.50%

23.90% 27.12% 63.75% 28.02%

4.82% 22.74% 25.08% 0.00%

Probability Note 4 Note 5 0.22% 11.51% 6.04% 0.00%

and more similar to A notes of closely related chickadee species. In addition, we did not detect another note type (i.e., B note), that, albeit rare, has been previously documented in Mexican Chickadee calls (Ficken 1990) as well as the calls of other chickadee species (e.g., Black-capped Chickadees, Charrier et al. 2004). There are at least two possible explanations for this departure. One possibility is that our recordings did not occur in contexts in which Mexican Chickadees use B notes; the other is that B notes are sufficiently rare, or even absent, in the calls of this species. This may be further support for Mexican Chickadees belonging to the brownheaded clade of chickadees as the only other brown-headed chickadee for which the chick-a-dee call has been studied to date, the Chestnut-backed Chickadee, also lacks B notes (Hoeschele et al. 2009). As well as classifying and analyzing individual notes and note features, we examined the syntax of the Mexican Chickadee chick-a-dee call. There is a growing body of evidence suggesting that different notes within the chick-a-dee call convey different meanings, or are at least used in different contexts (e.g., Freeberg & Lucas 2002, Mahurin & Freeberg 2009). Based on previous field research with Mexican Chickadees, the composition of any given chick-a-dee call appears to depend on the context in which that call is given (e.g., Ficken et al. 1994). Hailman et al. (1987) proposed that repetitions of note types within Black-capped Chickadee calls may serve to convey the level of

0.66% 7.95% 1.51% 0.48%

Total # Calls Note 6

Note 7

Note 8

0.00% 5.75% 2.72% 0.00%

0.00% 4.38% 0.30% 0.00%

0.00% 1.64% 0.00% 0.00%

324 107 261 205

motivation or urgency of the signaller. Ficken (2000) found that when Mexican Chickadees were approaching a disturbing stimulus, such as a predator, they produced more A notes, likely indicating alarm. Because of these findings, we analyzed call syntax using two methods: an expanded syntax method and a condensed syntax method (see Methods above). The condensed method may give a more accurate picture of the repertoire of the Mexican Chickadee chick-a-dee call while the expanded method may highlight the level of motivation or urgency of the caller. Call syntax appears to encode information not only based on the number of repetitions of note types in a call, but also based on the types of notes produced within a call. Prior research has shown that Black-capped Chickadees use syntax information for species identification in the field (Charrier & Sturdy 2005). Having a fixed set of syntactical rules within a species may be important for efficient message production and perception. Our sample of Mexican Chickadee calls follows a consistent rule of note-type production, in common with other chickadee species studied to date (Bloomfield et al. 2004, 2005, Charrier et al. 2004, Hoeschele et al. 2009). One caveat to our findings is that our samples were all recorded from birds in Portal, Arizona. This location comprises an isolated population in the northern-most range of the Mexican Chickadee. Gammon & Baker (2004) found that an isolated population of Black-capped Chickadees in Fort Collins, Colorado produced a set of song

Mexican Chickadee chick-a-dee call

types not found in other Black-capped Chickadees across North America. It has also been shown in the Chestnut-backed Chickadee that vocalizations may differ slightly in different geographic regions of a species’ typical range. As geographic distance increased between populations of Chestnut-backed Chickadees, differences in vocalizations also increased regardless of whether or not populations were isolated (Hoeschele et al. 2009). Thus, the isolated population of Mexican Chickadees that we sampled may produce vocalizations dissimilar from conspecifics in other regions. A more extensive sample of Mexican Chickadee calls taken from many areas, as well as focused recordings taken from the same individuals over time in varying contexts, would allow us to investigate if there are various dialects in the call of the Mexican Chickadee. Taken together, our studies provide a detailed description of the Mexican Chickadee chick-a-dee call. We have identified, described, and classified the note types present in our sample of this call, quantitatively measured features of those note types, and examined syntax rules in this call. Studies such as these provide a necessary foundation for future field and laboratory experiments aimed at understanding the messages encoded in this call as well as the mechanisms the birds may be using to decode those messages. Also, because this work has followed similar methodology to bioacoustic studies of other chickadee species’ chick-a-dee calls, it provides a foundation for future in-depth comparative studies that will examine the similarities and differences among the chick-adee calls of the Poecile genus.

ACKNOWLEDGEMENTS This research was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) discovery grant, Canada Foundation for Innovation (CFI) New Opportunities and Infrastructure Operating Fund grant, start-up and CFI partner funding from the University of Alberta, and an Alberta Ingenuity New Faculty Grant to CBS. MKM was supported by an NSERC CGS-M and an Alberta Ingenuity Studentship. MH was supported by an NSERC PGS-D, Alberta Ingenuity Studentship and a Queen Elizabeth II Master’s Scholarship (QEII) from the University of Alberta. We warmly thank Millicent Ficken for providing her extensive field recordings of Mexican Chickadees and for her pioneering research on

159

Mexican Chickadee vocalizations; the Borror Laboratory of Bioacoustics (Department of Evolution, Ecology, and Organismal Biology, Ohio State University, Columbus), and the Macaulay Library of Natural Sounds (Cornell Laboratory of Ornithology, Ithaca, New York) for their recordings and Scott W. J. Robson for stimulus preparation. REFERENCES Baker M. C., Becker A. M. 2002. Mobbing calls of black-capped chickadees: effects of urgency on call production. Wilson Bull. 114: 510–516. Betz N. E. 1987. Use of discriminant analysis in counseling psychology research. J. Couns. Psychol. 34: 393–403. Bloomfield L. L., Charrier I., Sturdy C. B. 2004. Note types and coding in parid vocalizations. II: The chick-a-dee call of the mountain chickadee (Poecile gambeli). Can. J. Zool. 82: 780–793. Bloomfield L. L., Phillmore L. L., Weisman R. G., Sturdy C. B. 2005. Note types and coding in parid vocalizations. III: The chick-a-dee call of the Carolina chickadee (Poecile carolinensis). Can. J. Zool. 83: 820–833. Burt J. 2006. Syrinx Version 2.6h [computer program]. www.syrinxpc.com. Accessed 30 June 2009 Charrier I., Bloomfield L. L., Sturdy C. B. 2004. Note types and coding in parid vocalizations. I: The chick-a-dee call of the black-capped chickadee (Poecile atricapillus). Can. J. Zool. 82: 769–779. Charrier I., Lee T. T-Y., Bloomfield L. L., Sturdy C. B. 2005. Acoustic mechanisms of note-type perception in blackcapped chickadee (Poecile atricapillus) calls. J. Comp. Psychol. 199: 371–380. Charrier I., Sturdy C. B. 2005. Call-based species recognition in black-capped chickadees. Behav. Process. 70: 271–281. Dixon K. L., Martin D. J. 1979. Notes on the vocalizations of the Mexican chickadee. Condor 81: 421–423. Doupe A. J., Kuhl P. K. 1999. Birdsong and human speech: common themes and mechanisms. Annu. Rev. Neurosci. 22: 567–631. Engineering Design 2008. SIGNAL Version 5.10.25 [computer program]. Engineering Design, Belmont, Mass. Ficken M. S. 1990. Vocal repertoire of the Mexican chickadee 1. Calls. J. Field Ornithol. 61: 380–387. Ficken M. S. 2000. Call similarities among mixed species flock associates. Southwest. Nat. 45: 154–158. Ficken M. S., Hailman E. D., Hailman J. P. 1994. The chick-a-dee call system of the Mexican chickadee. Condor 96: 70–82. Freeberg T. M. 2008. Complexity in the chick-a-dee call of Carolina chickadees (Poecile carolinensis): Associations of context and signaller behaviour to call structure. Auk 124: 896–907. Freeberg T. M., Lucas J. R. 2002. Receivers respond differently to chick-a-dee calls varying in note composition in Carolina chickadees (Poecile carolinensis). Anim. Behav. 63: 837–845. Gammon D. E., Baker M. C. 2004. Song repertoire evolution and acoustic divergence in a population of black-capped chickadees, Poecile atricapillus. Anim Behav. 68: 903–913. Gill F. B., Slikas B., Sheldon F. H. 2005. Phylogeny of titmice (Paridae): II. Species relationships based on sequences of the mitochondrial cytochrome-β gene. Auk 122: 121–143. Hailman J. P., Ficken M. S. 1996. Comparative analysis of repertoires, with reference to chickadees. In: Kroodsma D. E., Miller E. H. (eds). Ecology and evolution of acoustic com-

160

M. Moscicki et al.

munication in birds. Cornell University Press, New York, pp. 136–159. Hailman J. P., Ficken M. S., Ficken R. W. 1985. The “chick-a-dee” calls of Parus atricapillus: a recombinant system of animal communication compared with written English. Semiotica 56: 191–224. Hailman J. P., Ficken M. S., Ficken R. W. 1987. Constraints on the structure of combinatorial “chick-a-dee” calls. Ethology 75: 62–80. Hoeschele M., Gammon D. E., Moscicki M. K., Sturdy C. B. 2009. Note types and coding in parid vocalizations: The chick-a-dee call of the chestnut-backed chickadee (Poecile rufescens). J. Acoust. Soc. Am. 126: 2088–2099. Lengagne T., Lauga J., Jouventin P. 1998. A method of independent time and frequency decomposition of bioacoustic signals: inter-individual recognition in four species of penguins. C. R. Acad. Sci. Paris 320: 885–891. Lucas J. R., Freeberg T. M. 2007. ‘Information’ and the chick-adee call: Communicating with a complex vocal system. In: Otter K. A. (ed). Ecology and behaviour of chickadees and titmice: An integrated approach. Oxford University Press, Oxford, pp. 199–213. Mahurin E. J., Freeberg T. M. 2009. Chick-a-dee call variation in Carolina chickadees and recruiting flockmates to food. Behav. Ecol. 20: 111–116. Nowicki S. 1989. Vocal plasticity in captive black-capped chickadees: the acoustic basis and rate of call convergence. Anim. Behav. 37: 64–73. Slater P. J. B. 2003. Fifty years of bird song research: a case study in animal behaviour. Anim. Behav. 65: 633–639. Smith S. M. 1991. The black-capped chickadee: behavioural ecology and natural history. Cornell Univ. Press, New York. SPSS Inc. 2006 SPSS® Version 15.0.0. SPSS Inc. Chicago. Sturdy C. B., Bloomfield L. L., Charrier I., Lee T. T-Y. 2007. Chickadee vocal production and perception: An integrative approach to understanding acoustic communication. In: Otter K. A. (ed). Ecology and behaviour of chickadees and titmice: An integrated approach. Oxford University Press, Oxford, pp. 153–166. Sturdy C. B., Phillmore L. S., Weisman R. G. 2000. Call-note discriminations in black-capped chickadees (Poecile atricapillus). J. Comp. Psychol. 114: 357–364. Sturdy C. B., Phillmore L. S., Price J. L., Weisman R. G. 1999. Song-note discriminations in zebra finches (Taeniopygia guttata): Categories and pseudocategories. J. Comp. Psychol. 113: 204–212. Templeton C. N., Greene E., Davis K. 2005. Allometry of alarm calls: Black-capped chickadees encode information about predator size. Science 308: 1934–1937. Therneau T. M., Atkinson B. 2010. Recursive Partitioning. Package ‘rpart’ v. 3.1–46.

STRESZCZENIE [Typy sylab i kodowanie w wokalizacji sikor: głos chick-a-dee u sikory meksykańskiej] Zrozumienie funkcji komunikacyjnych każdej wokalizacji wymaga wcześniejszego opisania i klasyfikacji podstawowych elementów, które ją tworzą. Głosy chick-a-dee u sikor z rodzaju Poecile stały się dogodnym modelem do badania związków między strukturą i składnią sygnału a jego znaczeniem, ponieważ składają się z sekwencji

dyskretnych elementów (sylab), które mogą być zestawiane w sekwencji w dość dowolny sposób, przez co u różnych gatunków z tego rodzaju kodują odmienną informację. Obiektem badań była sikora meksykańska, której głos chick-a-dee nie był do tej pory szczegółowo badany w oparciu o większy materiał. Celem pracy było: 1. zidentyfikowanie, opisanie i sklasyfikowanie sylab występujących w głosach sikory meksykańskiej analogicznie do tego jak zrobiono to dla innych gatunków z rodzaju Poecile; 2. przeprowadzenie ilościowej analizy parametrów częstotliwości i czasu zidentyfikowanych sylab; 3. przeprowadzenie analizy składniowej głosów. Analizowany materiał dźwiękowy pochodził z trzech źródeł, przy czym zawsze używano jedynie nagrań o wysokiej jakości oraz takich, których opis nie pozostawiał wątpliwości co do tego, że głosy pochodziły od pojedynczego osobnika. Ogółem materiał stanowiły 543 głosy złożone z 1360 sylab i wszystkie nagrania pochodziły z okolice Portal (Arizona, USA). Głosy i sylaby je tworzące poddano szczegółowym pomiarom w programie SIGNAL. Następnie pomiarów tych użyto do opisu zmienności sylab oraz określenia potencjału z jakim dana charakterystyka sylaby umożliwia jej kategoryzację do określonego typu. Stwierdzono wydawanie przez sikory meksykańskie czterech typów sylab opisanych odpowiednio jako: A, C, D, Dh, stosując skróty analogicznie do używanych wcześniej u innych gatunków Poecile (Fig. 1). Sylaby A były modulowanymi gwizdami o opadającej częstotliwości. Sylaby C miały charakter poliharmonicznych, krótkich gwizdów o wzrastającej częstotliwości. Sylaby D były szumami o zmiennej długości, a sylaby Dh stanowiły swego rodzaju hybrydę głosu A oraz D i rozpoczynały się modulowanym wysokim gwizdem, który następnie przechodził w niższy szum. Zmierzone parametry sylab (Fig. 2, Tab. 1) pozwalały na bardzo precyzyjną dyskryminację między ich typami i to przy zastosowaniu zaledwie trzech prostych zasad (Fig. 3). Sikory wydawały sylaby wewnątrz głosu chick-a-dee z dość stałą składnią: A→C→Dh→D, niemniej każda z sylab w obrębie głosu mogła z niego „wypadać” bądź czasami była zastępowana przez inne typy (Tab. 2, 3, 4). W porównaniu do innych sikor z rodzaju Poecile głosy sikory meksykańskiej są dość krótkie, jedynie ok. 35% głosów zbudowanych było z trzech lub więcej sylab. Nie stwierdzono występowania w sekwencji głosu chick-a-dee sylaby B co jest typowe dla innych sikor należących do tego samego kladu.