The social brain hypothesis postulates that social complexity promotes the evolution of social cognition. In particular the ubiquity of public information in social environments is expected to prompt the evolution of social learning. However, it is still an open question if social learning relies on a newly evolved learning module that deals with this particular type of learning, or if on the contrary the mechanisms underlying social learning are the same as those involved in general purpose learning. It has also been proposed that the difference may rely not on the learning mechanism itself, but rather on attentional processes that make social information available (e.g. eavesdropping). In this study we address these two questions by: (1) Characterising the behavioural, neural and molecular mechanisms underlying a social and an equivalent asocial learning task in zebrafish; for this purpose an observational fear conditioning paradigm was contrasted with a classical fear conditioning paradigm. (2) Developing a task to study social attention in zebrafish; for this purpose fish observed a pair of non-interacting conspecifics, a pair of interacting conspecifics or an empty tank, and their engagement with the stimuli was measured using a homemade video-tracking system. Behavioural data show similar learning rates in the two learning tasks but higher attention rates towards interacting conspecifics. Brain transcriptome data show differential expression of a small number of genes in the attention task, which are candidates to be involved in attention in zebrafish. These results will be discussed in the scope of the brain modularity hypothesis for social information.