Among the multiple ways to learn behaviors, from classical conditioning to insight learning, observational learning is among the most powerful learning strategies. It allows animals, from invertebrate to primate, to acquire new behaviors through observation of others’ actions.
Observational learning, which is a highly cognitive learning, may help to (i) save energy by avoiding learning via many cycles of trials and errors, which require a lot of metabolic energy, (ii) decrease the time of exposure to the predators, (iii) decrease the amount of learning that is needed, by adopting behaviors that are already proven useful by other individuals.
Learning by observation has been shown to be crucial in many types of adaptive behavior, like foraging, predator avoidance, mating decisions, fear learning, and problem-solving strategies in humans. It has been shown that the ability to learn by observation is already present at birth and plays a crucial role in developing and mastering the lingual skills, social interactions, or the use of various tools relevant to everyday life. More recently, some disorders such as the autism spectrum or dyslexia have been partially characterized by the inability to learn by observation.
However, among various strategies of observational learning, little is known about any beneficial effects that observation could have on a subsequent performance during navigation in spatial orientation tasks and to what extent it is possible for animals to acquire information about spatial characteristics of an environment solely from observing actions of their conspecifics navigating through that environment. Moreover and importantly, there have been ambiguous reports concerning a role of observation of errors, namely whether it is more beneficial for an observer to observe a perfect execution of a task or rather a learning process in which imperfect examples and their outcomes are presented.
In our study, we have addressed these issues using a set of procedures in which the spatial memory of adult rats, Rattus norvegicus, was tested in an eight-arm radial maze.
Learning of spatial information by observing others
In our original procedure, rats were subject to a spatial discrimination task in an 8-arms radial maze. During 9 daily sessions of training, they were required to learn fixed positions of three arms with food portions on their distal ends. Animals were divided into groups of demonstrators, performing a task, and observers, which first observed demonstrators navigating in the maze from the observational platform and were later tested in a similar task in order to find how the observation influenced their ability to learn the task.
We found that groups of observers, having observed the learning process of the demonstrators, learned the task almost two times faster than the control animals, which have observed only an empty maze for the same amount of time. This suggests that rats were able to construct a cognitive map during the observation and later use it during processing of the task.
The positive effect of observation, however, was strictly dependent upon the observers’ prior knowledge about the environment. Indeed, before the start of any training, rats had been exposed to the maze during a habituation stage, in which they were allowed to get familiar with the environment (visit all the arms) and also gain some specific initial information about it (presence and location of the food).
It turned out that observers which, after habituation stage, were convinced that food is in every single arm of the maze could benefit only from observing the demonstrators making a lot of erroneous visits to non-baited arms but not from observation of demonstrators which were doing no errors and visiting only the three arms with food.
On the contrary, observers that had experienced an entirely empty maze during their habituation stage (no food in the arms), performed better only after observing perfectly trained demonstrators, that is such visiting exclusively the arms with food.
Therefore, these results suggest that it is not the observed errors themselves that are beneficial but rather an information that was new for the observers, comparing to the initial state of their knowledge. Eventually, rats were able to extract information that was new to them and use it later, in order to detect locations and configuration of specific arms within the maze.
Moreover, it seems that observers were constructing a cognitive map during observation and used it during the testing. Indeed, to have a beneficial effect of the observation, the configuration of arms with food which was observed needed to correspond specifically to the one which was experienced later during the actual exploration, implying that what was learned was indeed a spatial information about specific locations and their spatial relations.
Our results suggest that some form of a cognitive map of the environment is created during observation, which is later translated into a different set of coordinates when an animal finds itself in the same environment as its conspecific whose navigation process it has observed previously from a different perspective of the observational platform.
Presumably, such strategy might be utilized by wild-living animals when exploring new environments in larger groups, in order to increase the survival chances of the individual members of a group. A group might delegate one of its members to explore every novel route and gain information about that place based only on how it affected such individual so that there is no need for every single animal to visit each place itself. In the case that a route leads to the bedside of a predator or some other potential threat, it could minimize a risk that the group as a whole needs to take.
Strategy of this kind would resemble a one reported before in rats, in which one member of a group is delegated to first try some new type of food in order to see whether it is not harmful.
It would allow not only to gain information about places that are novel but also dynamically update information about places that had already been visited before but which may have undergone some changes since then.
These findings are described in the article entitled Observational learning of a spatial discrimination task by rats: learning from the mistakes of others? recently published in the journal Animal Behaviour. This work was conducted by TiazaBem and Bartosz Jura from the Polish Academy of Sciences, and Bruno Bontempi and Pierre Meyrand from the Université de Bordeaux and CNRS, Institut des Maladies Neurodégénératives.