Joint actions refers to any form of social interaction whereby two or more individuals coordinate their actions in space and time to bring about a change in the environment. At present, there are debates in existing literature about when children have the ability to perform adult-like joint actions because of the difficulty in identifying what mechanisms underlie the behaviors observed in children. In order to coordinate successfully, adults always represent tasks/actions of their own and partners’ in a functionally equivalent way, in which their own performance would be interfered. This interference phenomenon can be used as a index for adult-like joint action in children. The current study investigated whether young children owned the ability of co-representation which brought with interference in joint tasks. In experiment 1, 186 children aged 3~5 years old were chosen randomly to complete Same tasks or Different tasks with experimenter(partner). In the Same task condition, children and their partner had to press mouse button to respond to the same stimulus, while in the Different task condition, they had to respond to two different stimulus. If children co-represent their partner in this joint task, then there should be a cost to performance when the partner is following a different rule, compared with following the same rule. However, if children do not co-represent, the rule their partner is following would not cause interference. The results indicated that there were no significant difference between Same and Different task condition in 3 years old children, while children aged 4 and 5 years old performed worse in Different task than in Same condition. It indicated that 4 years old children might begin to experience interference in Different task. However, a potential alternative explanation for the difference between the two conditions was that participants in the Different task were confused about their task rules, as children heard two different instructions, while in the Same task participants heard the same instruction twice. In order to exclude the potential interpretation, a condition that co-representation of partner’s task might be beneficial, but where confusion due to instructions would be detrimental was designed. In experiment 2, children aged from 4 to 5 years old switched which stimulus to respond to half way through the game. If the children in the Different task were representing their partner’s task in the first half of the experiment, then switch costs should be reduced, as children already had the representation of the opposite task in mind before they switched to it. In contrast, in the Same task, children had no reason to represent the other’s task as both participants shared the same task, thus the opposite task is novel, therefore they experienced more switch costs. The results in pre-switch part replicated the results in experiment 1 and children had less switch costs in the Different task compared with the Same task condition. It meant that the interference children experienced in Different condition was really from co-representation of their own and partner’s task. In general, 3 years old children could not co-represent partner’s task and experience interference in joint task, while 4 and 5 years old children had the ability to co-represent their own and partner’s task in joint actions.