When performing joint actions, people modulate instrumental actions to provide additional information for a coactor (Pezzulo, Donnarumma, & Dindo, 2013). Similarly, demonstrators adjust instrumental actions to make them more informative for novice learners (Brand, Baldwin, & Ashburn, 2002). It is unknown whether the kinematic modulations performed to facilitate prediction in joint action coordination and the modulations performed to transmit information about the structure of novel actions are unique, or whether a general type of modulation can take on multiple functions. The present study therefore investigated whether there are unique kinematic markers for demonstration and for different types of joint action. In three experiments participants performed a virtual xylophone task, where they played simple xylophone melodies either alone, for a learner watching them, or together with another participant, while their movements were recorded. Participants increased movement amplitude during joint action and during demonstration. However, during joint action, participants modulated specific velocity parameters depending on whether their joint action partner knew or did not know the action sequence to be performed. The results demonstrate that there are specific kinematic cues to communicate the time and location of upcoming actions to a joint action partner but that there are no unique kinematic cues expressing the “pedagogical” intentions of a demonstrator.