Association for Behavior Analysis International

Dogs are well known for their superhuman olfaction and audition. Applying behavioral principles and techniques, we can explore the limits of dogs’ sense modalities. Perceptual limitations, however, are rarely a limiting factor when it comes to field or laboratory applications with dogs. Instead, the main challenge is to reliably and precisely bring an identifiable and useful behavior under the control of the relevant stimulus. For example, in a scent-detection scenario, an ideal training outcome is that the dog emits an “indication” response each time a target scent is encountered and never emits the response when the target is not encountered. Because dogs, like people, are never operating in the presence of a single stimulus but, instead, in the presence of a “stimulus soup,” it is also important to know whether stimuli operating through specific modalities are more or less likely to gain control over behavior than those operating through other modalities. In this series of presentations, the topic of stimulus control is explored in the context of basic and applied canine research. Although these presentations are related to canine behavior, researchers and practitioners in other areas may benefit from attending.

Practitioners and Applied Researchers

The development of differential stimulus control has long been of interest since the seminal paper by Reynolds (1961). The responses of dogs to a compound stimulus, composed of visual and auditory elements, was investigated. Twelve dogs were trained to perform a simple behaviour in response to a compound stimulus composed of a hand signal and a voice command. Then during test trials each modality was presented alone and the dogs’ responses were recorded. These test trials were interspersed with standard training trials of the compound stimulus. It was found that for eight dogs the voice command came to control the behaviour more than the hand signal, and for the remaining four dogs the hand signal controlled the behaviour more than the voice command. Each dog, therefore, showed a strong preference in responding to one modality over the other. The learning of the voice command overshadowed the learning of the hand signal for the majority of dogs. For the minority of dogs the learning of the hand signal overshadowed the learning of the voice command. The results suggest that individual dogs may show differences in their response to certain stimuli. These differences are discussed in relation to differences in prior experience and possibly to inherited characteristics of the dogs. The ways that the dogs in this study learned about the relevance of elements of compound stimuli has implications for the methods that are used in training dogs and these are discussed.

Scent-detection research and practice is often compromised by cuing and subjectivity. With cuing, stimuli other than the target scent gain control over behavior and can lead to erroneous experimental results or ineffective applications. Subjectivity in scent-detection research stems from application of topographical, rather than functional, response definitions. These issues can be resolved to some extent by training and testing under blind conditions, but such conditions can be difficult to arrange. We have been conducting research with an automated apparatus that, apart from an initial shaping period, requires no human involvement in the training or testing procedures and can accommodate a wide variety of samples. A brief description of this automated approach is provided, followed by a description of some results obtained from a series of research projects carried out using this approach. Basic research carried out with 5 dogs focusing on the indication response itself is highlighted, as this was demonstrated to be a critical factor in determining hit rate and false indication rate. Outcomes from an applied research project carried out with another 5 dogs aimed at evaluating dogs’ ability to identify water samples that have contained a target fish species (koi carp) are also summarized.

Complex odor mixtures are thought to be perceived configurally, implying that there is little identification of the individual components in the stimulus mixture. Prior research has suggested that the chemical and or perceptual similarity of components in a mixture may influence whether they can be detected individually; however, how training influences the identification of individual components in odor mixtures is less clear. Identification of individual odorants is critical for dogs tasked with discriminating between Home Made Explosives and very similar, but innocuous, complex odor mixtures. In a cross-over experimental design, we evaluated the effect of two training procedures on dogs’ ability to identify the presence of a critical oxidizer (i.e. component in explosives) in complex odor mixtures. In the “Mixture training” procedure, dogs received trial variable odor mixtures with and without an oxidizer. In the more typical procedure for canine detection training, dogs were presented with the pure oxidizer only, and had to discriminate this from decoy mixtures (“target-only” training). Mixture training led to above chance discrimination of the oxidizer from variable backgrounds and dogs were able to readily generalize performance, with no decrement, to mixtures containing novel odorants. Target-only training, however, led to a precipitous drop in hit rate when the oxidizer was presented in a mixture background containing either familiar or novel odorants. Furthermore, by giving Mixture training to dogs previously trained with the target in isolation, they learned to identify the oxidizer in mixtures very quickly. Together, these results demonstrate that the training method has significant impacts on the perception of components in odor mixtures and highlights the importance of olfactory learning for the effective detection of Home Made Explosives by dogs.