Bottlenose dolphins have evolved individually distinctive signature whistles that they invent in the first few months of their lives and then use to broadcast their identity to others.
In many ways these whistles are similar to human names. However, humans not only use names to recognize others, they can also use voice cues: features that make our voices distinctive and recognizable to those who know us.
Over the last few years we have conducted playback experiments during capture-release sessions to investigate whether dolphins are able to recognize each other by voice as well as by signature whistles.
Analysis of these data showed little evidence for voice recognition, but the variation in the reaction of animals was much larger than in our previous experiments. Such variation means that more data are needed to identify whether or not there is an effect. We therefore used the capture-release sessions this year to do more voice recognition playbacks. We will continue these through next year and then re-analyze our sample to get a more definitive result regarding whether or not voice recognition occurs in dolphins.
Although these experiments have not yet yielded the result we were trying to address, we found some exciting preliminary results in analysis of the acoustic data. Several males copied the playback stimuli, with some even replicating noisy artifacts in the stimuli. In addition, significantly more male (6/12) than female (0/14) target animals produced a distinctive whistle contour (which we called the “M whistle”) in response to playbacks of a variety of different non-signature whistle stimuli (see spectrograms – plots of frequency vs. time – below). While dolphins have been found to produce context-specific pulsed sounds such as brays and pops, this is the first evidence for a context specific non-signature whistle vocalization. Claire Stuhlmann of the University of Pennsylvania, who worked with us on this project this summer, focused on looking for occurrences of this whistle in contexts other than playbacks of non-signatures; her work is ongoing. Given the evidence for vocal learning of whistles in dolphins, it is likely that this “M whistle” is a learned signal. These playbacks provide some of the first insights into the role of non-signature whistles in the natural communication system of dolphins.
Finally, in a new project, we looked at the parameters that encode identity in signature whistles. We know that the modulation pattern of signature whistles carries this information, but we do not know the minimum information needed to ensure recognition. Together with Arik Kershenbaum from the National Institute of Mathematical and Biological Synthesis in Knoxville, TN, we therefore analyzed signature whistles with an algorithm commonly used to identify music tunes. This method only looks at when the frequency changes in a whistle and in what direction (to become higher or lower pitched). We found that this algorithm can be used to recognize signature whistles but that it does not perform as well as visual classification of spectrograms (see below) by humans. It is too early to say how useful this method might be; we will be testing it to see how well it works with additional data sets.
This article was published on page 11 in the January 2014 Nicks n Notches.