Multi-finger configurations: a paradigm to study learning novel muscle synergies

Learning novel patterns of muscle activity, such as when producing a new guitar chord, is an essential aspect of human motor skill learning. However, most existing motor skill learning paradigms do not focus on this aspect, but rather deal with modification or combination of previously well-practiced actions (e.g. learn sequence of single-finger presses, adaptation of reaching movements). The study presented in this thesis aims to develop and validate a new well-controlled lab task to study learning novel muscle activity patterns. Participants practiced on 242 unique multi-finger configurations, involving flexion and extension at the metacarpophalangeal joint. On the first day, some of the configurations were so difficult that participants could not produce them. Nonetheless, after 3 days of practice, they were able produce all 242 combinations. We then quantified the difficulty of each chord across participants, and compared different explanatory models, including the cognitive complexity of the chord, the force directions, and the patterns of muscle activity. The model based on patterns of muscle activity provided the best predictive accuracy, indicating that some characteristic of the specific muscle activity pattern makes chords difficult to produce. To investigate the nature of these characteristics, we quantified the naturalness of these muscle activity patterns by comparing them with the muscle activity patterns that occur during natural everyday actions. We found that the probability of a given pattern occurring during natural actions, as well as the absolute amount of muscle activity, determines the difficulty of producing novel muscle activity patterns.