Collecting data on dancers' bodies and their technical performance has now reached fairly sophisticated levels with complex laboratory three- dimensional filming equipment and motion analysis software along with EMG recordings. However, examining and documenting the process of teaching and learning dance skills in a realistic sense still requires observation/filming of class work followed by the tedious process of coding by trained professionals to document the skill acquisition process. This provides some unique challenges of protocol and methodology to ensure that the data collected in this way meets the rigorous standards of reliability and reproducibility to pass the judgement of an editorial review process and achieve academic credibility. One is also limited in what type of research questions can and cannot be answered through observational coding. Several examples from the area of motor learning research will be used to illustrate the challenges of developing a good coding system. One area is the use of live versus video coding in evaluating performance. Another problem is trying to determine what is a dance ‘skill' and what components of this skill can we evaluate? For example, is it realistic to attempt to ‘score' flow or musicality? Even if we limit ourselves to the motor actions themselves, what is a realistic detail that can be reliably identified in terms of technical performance that may be required for a pre and post test learning experiment? A third area is to look at the learning of dance sequences. Here we face a problem of what unit of analysis we should code in a sequence and whether learning includes only the mechanical performance of physical actions or other aspects. Illustrations from two research projects will address some of the challenges of designing good coding systems for dance. Keyword(s): dance, research