The evolutionary history between humans and dogs is unique: the emergence of domestic dogs is temporally contiguous with the first human settlements approximately 12,000 years ago (although it should be noted that there is some evidence that domestic dogs may have diverged from wolves much, much earlier than this; e.g. Vila et al., 1997). In the ecological niche of human society, dogs have thrived both as working partners and as companions. There are thought be over 400 breeds of domestic dogs worldwide, showing unique intraspecific variations in morphology and behaviour. No other animals has such a diverse range of breeds and types.
As a species with great socio-cognitive ability, dogs have evolved several modalities for communication. Body language or "postural signalling" forms a very important part of the dog's communicative world - these signals are common across all dog breeds, and more generally across many different species of canids (e.g. wolves, coyotes, etc). Some researchers believe that no other form of communication is necessary in dogs, as postural signals enable the expression of any important information. However, this would imply that all vocalisations are essentially meaningless, which is clearly not the case.
Recent studies have investigated the role of barking in dogs and identified that barks are a graded vocalisation that have the potential to express a range of information, such as affective or motivational state and individual identity (Yin, 2002; Yin & McCowan, 2004; Pongracz et al., 2005; 2006; Molnar et al., 2006). Furthermore, it has been shown that barks recorded in different situations can be classified into accurate subsets by statistical analyses and automated algorithms (e.g. Yin, 2002; Molnar et al., 2008). Barking thus clearly holds functional value, and "meaningless" barking is likely to be a behavioural abnormality or stereotypy.
It is not, however, just a question of information about mood or motive. In the field of vocal communication, it is well understood that animals transmit in their calls additional information about themselves, for example age, sex or body size. These are potentially vital characteristics for an animal's chances at reproduction and survival, and are encoded by acoustic variables in the call. Specifically, much interest has been given to fundamental frequency and formant dispersion.
In the Vocal Dog project, we are interested in investigating whether dog vocalisations have the potential to transmit information about the caller. Uniquely (and enabled by the dog's co-evolutionary history with humans, we are looking at interspecific interactions as well as taking the more conventional conspecific approach. We are thus investigating dog vocal communication within two frameworks:
(1) Size communication
To understand perception and attribution of size information in growls, we resynthesise the formants of growls in order to simulate larger or smaller dogs (so that the growl is acoustically identical for all dimensions other than formants) and use these in playback experiments. We are interested in finding out how well humans, other dogs and prey species (e.g. sheep) attend to formants, and whether they can attribute size information on the basis of formants.
To date we have confirmed the correlation between formant dispersion and dog body size (previously demonstrated by Riede & Fitch, 1999) and shown that humans are able to use formants to accurately judge the size of domestic dogs. We also found that humans paid attention to the fundamental frequency of growls when making size assessments; however their reliance on fundamental frequency was only minor compared to their reliance on formants.
This research is published: Taylor, A.M., Reby, D., McComb, K. (2008). Human listeners attend to size information in domestic dog growls, Journal of the Acoustical Society of America, 123, 2903-2909. CLICK FOR PDF
(2) Context information
The term "context" is used here as any variable that identifies the situation in which a growl is emitted. We are thus interested in identifying acoustic dimensions that encode context in a predictable manner.
Go to the reference page for further reading.