In a large variety of contexts, it is essential to use the available information to extract patterns and behave accordingly. When it comes to social interactions for instance, the information gathered about interaction partners across multiple encounters (e.g., trustworthiness) is crucial in guiding one’s own behavior (e.g., approach the trustworthy and avoid the untrustworthy), a process akin to trial-by-trial learning. Building on associative learning and social cognition literatures, the present research adopts a domain-general approach to learning and explores whether the principles underlying associative learning also govern learning in social contexts. In particular, we examined whether overshadowing, a well-established cue-competition phenomenon, impacts learning of the cooperative behaviors of unfamiliar interaction partners. Across three experiments using an adaptation of the iterated Trust Game, we consistently observed a ‘social overshadowing’ effect, that is, a better learning about the cooperative tendencies of partners presented alone compared to those presented in a pair. This robust effect was not modulated by gender stereotypes or beliefs about the internal communication dynamics within a pair of partners. Drawing on these results, we argue that examining domain-general learning processes in social contexts is a useful approach to understanding human social cognition.In any domain of daily life and cognition, humans solve tasks and make decisions by using information that comes from multiple, different sources. It is quite obvious that we learn from previous experiences. We then use multiple sources of information to guide our behaviour in environments, make decisions about what is beneficial for us, and act in social situations (attributions, imitation). Most times however, not all information in the environment is useful. For example, if we eat fish and chips and later become ill, it is difficult to know which of the two made us ill, and people tend to select one or the other based on quantity. In fact, humans are particularly adept at selecting and learning from those sources which provide information about relevant outcomes. Hence, the idea of competition between different sources of information has been prominent in theories of learning. A wealth of data in the social sciences and psychology supports this assertion. Yet, the finding that multiple sources of information compete during learning is not ubiquitous. In some fields (i.e., spatial learning, category learning) and in experiments using animals, researchers have found facilitation (the opposite of competition) between multiple sources of information, and this has led to the development of specific theories that explain those findings, by assuming they are "exceptions". Thus, there are theories aimed at explaining competition, and theories aimed at explaining facilitation, but no consensus regarding the circumstances that lead to either of these opposite outcomes. Based predominantly on experiments conducted in nonhuman animals, Urcelay has recently hypothesized that competition and facilitation are two extremes of a continuum which is determined by multiple variables. The objective of this proposal is to test whether contiguity (that is, the temporal and spatial separation of events) is a critical determinant of these opposite findings. We predict that competition and facilitation are phenomena that can be observed across different tasks and domains of cognition. Therefore, we will test this general prediction across different learning preparations with increasing complexity. First, we wish to determine whether competition and facilitation are observed as a function of temporal separation between hypothetical food consumption and sickness in a predictive scenario such as the food-disease example mentioned above. The second objective is to extend these findings to an action-outcome task. Action-outcome learning underlies the sense of agency, free will and responsibility. We will investigate whether competition or facilitation are observed in a task in which participants' actions (i.e., press a button) and other environmental events (a signal) are presented simultaneously during learning. We will do this whilst manipulating the time between action and the outcome. We predict that when outcomes are presented soon after the action, stimuli will compete with action-outcome learning, but that the opposite will occur when outcomes are delayed. This is relevant to human behaviour because most of our actions are followed by delayed outcomes (e.g., saving for retirement; preparing for a child's birth). The final objective of this proposal is to investigate this prediction in spatial learning, which integrates the above mentioned objectives in a more complex and ecologically valid setting, whilst extending the prediction to spatial separation. Notably, in spatial learning both competition and facilitation have been observed, but the exact reasons for these discrepant findings have not been elucidated. In all three objectives, we will further investigate the psychological mechanisms (whether participants process different bits of information as separate elements, or as a whole) underlying competition and facilitation in these three scenarios. These findings have important implications for theories of learning, and for education.

The data was collected whilst participants participated in the experiments via an online platform (Prolific). In each experiment, participants experienced different conditions and also allocated to different groups (mixed designs). The experiment was written and hosted in Gorilla for online data collection. Participants were recruited through Prolific.