While most conversation analysts still reproduce today what we call the adjacency pair dogma, that is, the belief according to which conversations are basically organized under the form of pairs (e,g., question/response, invitation/acceptance, greetings/greetings, etc.), we show that this position is misguided and that conversation should actually be considered as organized under the form of triads. Even if conversation analysts acknowledge that a conversation is not only made of what is being said but more generally what is being done, they tend to forget this when the time comes to analyze the basic structure of conversations. This explains why they tend to ignore the third turn, a third turn that allows conversationalists to close a sequence. We illustrate this position by studying how street hypnosis sessions are organized.