Learning In Dogs : Social Learning & Motivation

Social learning

Social or observational learning refers to any form of learning where changes in behaviour are dependent upon or facilitated by the presence of other animals. Thorpe (1963) proposes three categories of social learning: (i) social facilitation, in which the behaviour of one animal prompts the same behaviour by another animal but the behaviour already exists in the second animal’s repertoire; (ii) local enhancement, in which the behaviour of one animal draws the attention of another animal to the specific place in the environment or the particular stimulus involved, but the second animal still acquires the new behaviour by trial and error; and (iii) true imitation, in which the behaviour of one animal prompts the exact same form of the behaviour by another animal and the behaviour is highly improbable in the second animal without this opportunity for observation.

Despite a lack of research on social learning in dogs, there is anecdotal evidence that dogs are capable of both social facilitation and local enhancement. An example of social facilitation is where a dog, reluctant to negotiate a set of slippery stairs, easily follows another dog up the stairs and, after that exposure, is comfortable traversing the stairs on its own. Trainers sometimes utilize social facilitation to encourage a dog to try a novel treat, to enter a new environment or to investigate a frightening object. An example of local enhancement is a dog that has never manipulated a food toy, like a Buster Cube™. After observing a demonstrator dog obtaining food from the toy, the second dog may learn a strategy for getting the food from the toy more quickly than a dog that was not given the opportunity to observe a demonstrator dog. Local enhancement is very likely the mechanism by which puppies in the Slabbert and Rasa (1997) study learned an odour-detection task more quickly after observing their dams performing the task than did control puppies that did not observe their dams at work.

True imitation can be concluded only after controlling for learning by social facilitation and local enhancement. Few, if any, convincing examples of imitation exist in species outside humans and the great apes (Whiten and Ham, 1992) and to my knowledge, there are no demonstrations of true imitative learning in dogs.

Motivation

In order to effect behavioural change through operant conditioning, there must be a motivating force in the form of a behavioural consequence. As noted above, behavioural consequences involve delivering or removing pleasant (appetitive) stimuli or delivering or removing unpleasant (aversive) stimuli. There are two classifications of such stimuli: unconditioned (or primary) and conditioned (secondary). Unconditioned stimuli require no prior learning to appreciate, such as food, water, social contact, sex and various noxious or painful events. For ease of administration, most dog training utilizes food and social contact as forms of unconditioned appetitive reinforcers, with social contact consisting of vocalizations (praise), physical touch (stroking) and interactive play (wrestling, chase, tug and retrieving games). Not surprisingly, the value of these typical rewards tends to vary amongst dogs; some dogs find little reinforcing value in social contact, whereas other dogs prefer interactive play to food. One dog finds retrieving especially motivating, while another dog enjoys tug. A study in which dogs were required to press a lever for food rewards revealed that one subject would press the lever approximately 30–40 times per session for standard dry kibble rewards, but would press the same lever well over 100 times per session for small tidbits of liver (Reid, unpublished). In general, animals perform faster and/or longer for larger or more palatable rewards (Mackintosh, 1974).

/or longer for larger or more palatable rewards (Mackintosh, 1974). Unconditioned aversive stimuli are usually vocalizations (reprimands), physical touch (restraint or other forms of ‘corporal punishment’ like hitting) or pain delivered through a training collar, such as choke or pinch collars that can be constricted around the dog’s neck or electronic collars that can be triggered to present a noxious spray (citronella) under the dog’s face or an electric shock to the dog’s neck. Beerda et al. (1998) demonstrated that dogs show more fearful postures and experience elevated cortisol levels (indicative of higher stress) in response to unpredictable aversive events, such as loud noises, falling objects and electric shock, than to predictable events, like an umbrella opening, a spray of water from a gun or physical restraint. This is consistent with findings from controlled laboratory examinations of other species: the ability to predict an aversive event can dampen its unpleasantness somewhat (Lockard, 1963).

Operant conditioning is especially sensitive to the temporal relation between behaviour and its consequence (Mackintosh, 1974). In some situations, learning can be seriously hampered with as little as a 0.5 s delay between the behaviour and the outcome (Grice, 1948). In other cases, learning is possible with delays as long as several minutes (Lett, 1973). The critical feature appears to be the occurrence of competing behaviours during the delay (Revusky, 1971). For instance, if the dog sits but then barks, turns its head and jumps up, all before the trainer is able to deliver the food treat, it is difficult for the dog to discern which behaviour led to the reward. While the timely delivery of food is quite feasible in an experimental chamber, it is quite another matter in the real world of dog training, especially when the dog may be working at a distance from the trainer

y when the dog may be working at a distance from the trainer. Fortunately, overcoming the problem of timing consequences is accomplished through the use of conditioned or secondary stimuli, which can be delivered immediately after the target behaviour. A conditioned reinforcer is a stimulus previously linked with an unconditioned reinforcer, such that it comes to predict the occurrence of the unconditioned reinforcement (Mackintosh, 1974). Pavlov’s metronome functioned as conditioned reinforcement for his dogs. Practical conditioned reinforcers tend to be acoustic, so that they can be delivered immediately after the behaviour, even at a distance. For many dogs, ‘Good dog!’ is a conditioned reinforcer because it reliably precedes other pleasant events, like treats and play. ‘Clicker training’, first described by Skinner (1951) and subsequently popularized by marine mammal trainer Karen Pryor (1984), involves the use of a metal noisemaker, called a clicker, as a conditioned reinforcer (see Fig. 8.3). The dog is taught that the sound of the clicker signals a tangible reward is on its way and, consequently, the click bridges the temporal delay between behaviour and reward (Williams, 1991). A related but distinct effect is that the sound of the click helps to mark the target behaviour and make it distinguishable from other behaviours (Lieberman et al., 1979); hence the clicker is often called a ‘marker’. A trainer can time the administration of a click much more precisely than the delivery of food treats or toys, thereby reducing uncertainty and speeding the learning process for the dog. Less systematic in dog training is the use of conditioned negative reinforcement, also known as a safety signal. Imagine a situation where the dog jumps up on the kitchen counter. The owner shouts at the dog and the dog jumps off. Were the dog to remain, the owner would positively punish the dog in some way. Instead, the owner praises the dog for jumping back down. This praise functions as a safety signal to the dog, signalling that it has successfully avoided the positive punishment (I think of this as the ‘Whew!’ effect), thereby negatively reinforcing the action of jumping off the counter. Trainers who use devices like electronic collars sometimes establish tones as audible safety signals to convey that the dog has performed the target behaviour and escaped or avoided the potential aversive stimulus.