Many argue that dogs show unique susceptibility to human communicative signals that make them suitable for being engaged in complex co-operation with humans. It has also been revealed that socially provided information is particularly effective in influencing the behaviour of dogs even when the human's action demonstration conveys inefficient or mistaken solution of task. It is unclear, however, how the communicative nature of the demonstration context and the presence of the human demonstrator affect the dogs' objectchoice behaviour in observational learning situations. In order to unfold the effects of these factors, 76 adult pet dogs could observe a communicative or a non-communicative demonstration in which the human retrieved a tennis ball from under an opaque container while manipulating another distant and obviously empty (transparent) one. Subjects were then allowed to choose either in the presence of the demonstrator or after she left the room. Results showed a significant main effect of the demonstration context (presence or absence of the human's communicative signals), and we also found some evidence for the responsemodifying effect of the presence of the human demonstrator during the dogs' choice. That is, dogs predominantly chose the baited container, but if the demonstration context was communicative and the human was present during the dogs' choice, subjects' tendency to select the baited container has been reduced. In agreement with the studies showing sensitivity to human's communicative signals in dogs, these findings point to a special form of social influence in observational learning situations when it comes to learning about causally opaque and less efficient (compared to what comes natural to the dog) action demonstrations.