• Migratory groups are large and may include those passing through Plettenberg Bay
6.4.1. Methodological considerations in the study of geographic and species variation
The core parameter values from signature whistle types identified using two different vigilance parameters (91 and 96) and contour presentation techniques gave similar results, generating confidence in the trends identified between populations and the robustness of the techniques. The whistle sampling method used to investigate variation (signature whistle types or contours) had a significant influence on the values for all core parameters, although not all populations were affected equally. Where differences existed in frequency values, the signature whistle types generally had the lowest frequencies of the 3 contour sampling methods. As signature whistles are likely to be used as contact calls, to maintain group cohesion (Janik and Slater, 1998, Watwood et al. 2005), selective preference for lower frequency components might be expected, as these transmit further in the marine environment (Janik 2000a). Conversely, the maximum frequency of signature whistle types from Harderwijk and Scotland West coast were greater than those from the other two contour sampling methods. It is possible that the inclusion of known signature whistles in the Harderwijk sample shifted the maximum frequency upwards, either because the quality of the recordings was particularly high so that high frequency components suffered no attenuation effects or because whistle frequency parameters modify under stressful circumstances (Esch et al. 2009a)such as the isolation contexts in which these recordings were made. However, Esch et al. (2009a) only detected an effect of captures in frequency parameters for mothers caught with dependent calves, and on the whole they do not appear to differ between capture –release and undisturbed conditions (Esch et al. 2009a). For
Scotland West Coast, the relationship is not easily explained. Whilst it might be a result of the small sample size of signature whistle types from this region, selective advantages of higher frequency SWTs which may be more easily localised at short distances (Esch et al. 2009a), cannot be discounted for this population.
For most populations (7/10) the duration and number of inflection points were much higher for signature whistle types, compared to the contour samples. This reflects the use of disconnected multiple-looped SWTs by most populations (Chapter 4) and demonstrates a source of potential bias in previously reported characteristics of bottlenose dolphin whistle features (e.g. Wang et al. 1995, Morisaka et al. 2005a, Hawkins, 2010). Investigations of geographic variation in whistles in which the authors do not join the loops of disconnected multi-loops together to form one vocal output (e.g. Morisaka et al. 2005a,b, Oswald et al. 2003, 2007), are likely to be overlooking some of the complexity encoded in the whistles. The process of joining loops used here, did so without accounting for the period of silence between successive loop emissions (inter-loop interval, ILI). Variation in the ILI is apparent at the individual (Esch et al. 2009b) and population level (Chapter 4), indicating that this might be further used to signify emotional state, individual or population variation (Esch et al. 2009a, Esch et al. 2009b). Also, if measures of signature whistle duration include the period of silence (c.f. Caldwell et al. 1990), then durational differences highlighted in this study would be even greater in populations using disconnected multi-looped whistles.
The prevalence of signature whistles in recordings of whistle vocalisations varies with behavioural context (Caldwell et al. 1990, Janik and Slater 1998, Watwood et al. 2004) group composition (Cook et al. 2004) and recording protocol (for example, line surveys sampling several groups, each for a short period of time vs. focal follows) and can range from 39% to 52% in wild animals. (Cook et al. 2004, Watwood et al. 2005). As has been suggested in avian acoustics (Mundinger 1979), pooling data across contours is an acceptable way to obtain basic structural information on the frequencies used, particularly if a wide variety of individuals and contexts are sampled. Given that signature whistles make up such a high proportion of the vocal repertoire, it is likely that whole-repertoire analyses will reflect the trends apparent in the signature whistle frequencies. For instance, individual loops of DCMLs will have the same or similar measures for many frequency parameters (start, end, minimum, maximum and mean frequency) as the entire DCML identified though SIGID. However, applying a method such as SIGID to identify functional call types is a necessary pre-requisite for detailed descriptions of whistle structure and complexity, which cannot be done without first identifying the units of whistle production, including disconnected multi-looped signature whistle types.It was not possible to separate whistle contours into signature and definite non-signature whistles in this analysis, as bottlenose dolphins can produce signature whistles as single occurrences and the SIGID method does not necessarily identify all signature whistles apparent in each data set (Janik et al. in press). Further investigations attempting this would be worthwhile, to better quantify systematic differences between whistles used to convey identity information and non-signature whistles.
The co-efficient of modulation is widely applied as a reliable measure of frequency modulation (McCowan and Reiss 1995b, Morisaka et al. 2005a, May-Collado and Wartzok 2008) and was used here for wider comparison. It was highly correlated with inflection points, supporting its use. However, there are some important methodological issues concerned with using this measure to accurately quantify modulation in whistles. The coefficient is calculated by summing the average change in frequency over 20 frequency points. However, using this calculation, a contour which increases linearly between each of the 20 frequency points, would receive a co- efficient of modulation score of 1.9. Conversely a whistle increasing by steps has a coefficient of modulation of 1 (see worked example, Appendix 13). Furthermore, a contour which decreases in frequency by the same amount that it increases, has the equivalent COFM of the straight contour, without accounting for the occurrence of an inflection point. Therefore, whilst the coefficient can measure degrees of modulation in highly modulated contours, this parameter does not accurately measure modulation in the manner most readers (and likely dolphins) perceive it to exist. As such, a more realistic and accurate measure of modulation is the number of inflection points, whilst the frequency gradient (change in frequency/duration) gives some indication of the overall steepness of contours. In many studies (e,g, Wang et al. 1995, Morisaka et al, 2005), whistle features such as inflection points, are visually assessed. However, standardised thresholds for parameters, such as those applied here, are more objective and should be used for any comparison of population level variation (c.f. Oswald, et al. 2003, 2007).
For any analysis of whistle frequency parameters it is important that the bandwidth of the recording equipment exceeds the maximum fundamental frequency of the whistles encountered. Without this, several parameters, notably maximum frequency, end frequency (Oswald et al. 2004) and whistle duration may be miscalculated and the overall shape of the frequency contour might be misrepresented. A small percentage of the high quality contours from Doubtful Sound and Scotland East Coast were bandwidth limited and not included in the overall comparison. It is therefore possible that the frequency parameters of the SWTs from these regions are underestimated. Indeed, full bandwidth recordings from Doubtful Sound report an maximum frequency of 41 kHz (Boisseau 2005). Therefore, the differences between T. aduncus and T. truncatus could be even greater than reported here. The highest mean maximum frequency reported for T. truncatus signature whistles is 27.3 kHz (Esch et al. 2009b).In this analysis, the signature whistles for Doubtful Sound and Florida had the highest mean maximum frequency (16.6 and 16.3 kHz respectively). However, several signature whistles identified had maximum frequencies as high as 32 kHz. Overall, the maximum frequency recorded for any contour was 35 kHz, from Scotland West Coast. The recording conditions, including the distance of the vocalising animal from the hydrophone and ambient noise are likely to influence whether the high frequency components of whistles are recorded. Given this, a recording sampling rate of 96000 Hz or above is strongly recommended and the highest quality recordings should be used in any investigation of geographic variation in the frequency characteristics of dolphin whistles.