Musical improvisation is commonly understood as an on-line compositional-performative activity, as a result of which a musical piece is created. Generative theories of music cognition modelled music improvisation as a musical process where proactive and retroactive constraints regulate on-line music production (Pressing, 1988; Johnson Laird, 2002). In this paper we re-frame musical improvisation and situate it into the wider context of the theory of embodied music cognition (Leman, 2008). Music improvisation is now understood as an activity built-in during the course of the improviser?s embodied interaction with the environment. Embodied music improvisation is the result of the unfoldying of action-reaction cycles, where the improviser plays, perceives, judges and make decisions based on the results of his on-course embodied action (Leman 2008). Embodied action includes the processing and recovering of distal and proximal cues emerging from the unfolding performance. As long as an action-reaction cycle takes place the improviser (i) develops certain expectations about the sonic result of his future action; (ii) perceives -both corporeally and aurally- the results of his ongoing action; (iii) actual embodied music performance constraints future responses and generates new expectations; and (iv) reaction may be guided by those expectations and can be more or less automatic and/or reflective. In this paper it is assumed that the improviser ?as the receptor of his own ongoing music production process - develops schematic (musical structure-dependent) and dynamic (ongoing events-dependent) expectations (Huron, 2006) related to the proactive and retroactive constraints that operate on-line, being these expectations part of the embodied activity that takes place in the action-reaction cycles described above. An experimental study was developped to test this assumption. 15 improvisers were required to perform the improvisation of a solo over a recorded base of a twelve-bar blues that was presented four times. After the first presentation of the original base, participants improvised over three modified versions containing rhythmic, harmonic and/or formal changes. Participant?s responses were video and audio-recorded. Responses were both transcribed and audio-analysed. Based on the results of a previous study (Perez, 2011) it was assumed that the produced changes would provide on-line coherent cues to prompt the emergence of both schematic and dinamic expectations that guide improvisers? responses. It is expected that experimental data ?analyses of results still in progress- provide significant information about the way expectations operate in the embodied activity that takes place during the action-perception cycles of musical improvisation.