The effect of task structure, practice schedule, and model type on the learning of relative and absolute timing by physical and observational practice

Three experiments compared learning of relative and absolute timing of a sequential key-pressing task by physical and observational practice. Experiment 1 compared a task with a complex internal structure (goal proportions of 22.2, 44.4, 33.4 on the three movement segments) to one with a simpler structure (goal proportions of 33.3, 33.3, 33.4). Observers only learned the relative timing as well as physical practicers when the internal structure was simple, but learned the absolute timing in both conditions. Experiment 2 compared variable (700, 900, and 1100 ms overall time) with constant practice (900 ms overall time). Observers of constant practice models learned the relative timing better than no-practice control participants, but not as well as the models, while observers of variable practice models learned the relative timing no better than the control group. Observers in both practice conditions were able to produce the absolute timing as well as those who physically practiced. In Experiment 3 observers of an expert model were able to produce the relative timing as well as those who physically practiced the skill, while those who observed learning models were not. All observers and the physical practice participants were able to produce the overall duration as well as the expert model. The results of these three experiments support earlier findings that increasing stability during practice promotes better learning of relative timing, but that absolute timing can be learned under less-stable conditions (Lai, Shea, Wulf, & Wright, 2000b). These findings also have important implications on the limitations of Scully and Newells' (1985) prediction that relative timing, but not absolute timing, could be learned by observation. Experiments 1-3 along with earlier findings (Black & Wright, 2000) have consistently found that absolute timing could be learned by observers even as the nature of the task, practice schedule, and model are manipulated. Furthermore, the results suggest a limitation to the effectiveness of learning models (Adams, 1986; McCullagh & Caird, 1990).