Time Scales in Motor Learning

运动技能的学习以行为的持续变化为特征。运动行为变化的指标有很多，行为变化发生的时间尺度也有很多（变化率和时间段）。这项研究的一个中心命题是，时间尺度是描述和预测我们推断为学习的行为变化的基础。该研究基于非线性动力系统的概念和工具的理论框架；这个框架的设计是为了解释运动技能学习传统上显示的持续（相对长期，如几天、几周、几个月）和短暂（相对短期，如一次又一次）的变化。过去100年对运动技能学习的研究表明，在学习曲线中揭示了许多不同的变化函数，如指数、幂律、s形、逻辑、突然的“不连续”。通常，不同的学习理论对用于拟合学习数据的数学方程中固有的变化时间尺度做出不同的假设。此外，学习理论倾向于关注学习的一种功能，而不是适应完整的学习功能。这项研究将检验一个命题，即来自非线性动力学的一小部分原理可以产生运动学习中观察到的所有标准变化函数。在动力系统框架内进行的一系列实验将检查，在一系列运动任务中，运动学习变化的时间尺度。这项研究将提供一个概念的第一个测试，一个统一的和简约的动态描述的时间尺度的变化，可以推导出一套既定的短期和长期的学习功能在运动学习。这种对运动学习时间尺度的理论和实验方法的前景是，它将带领我们超越传统的学习描述，走向人类运动学习的预测科学，从理论上将神经网络方法与人类认知和人工系统学习联系起来。

The learning of movement skills is characterized by persistent change in behavior over time. There are many indices of change in motor behavior and many time scales (rates of change and time periods) over which the change in behavior occurs. A central proposition of this research is that time scales are fundamental in the description and prediction of the change in behavior that we infer as learning. The research stems from a theoretical framework based upon the concepts and tools of nonlinear dynamical systems; this framework was devised to account for both the persistent (relatively long-term, such as days, weeks, months) and transitory (relatively short-term, such as trial to trial) changes traditionally shown for the learning of motor skills. The last 100 years of research on the learning of motor skills has shown that a number of different functions of change are revealed in learning curves, such as an exponential, power law, S-shaped, logistic, sudden "discontinuous". Typically, different theories of learning make different assumptions about the time scales of change that are inherent in the mathematical equations used to fit learning data. Moreover, theories of learning tend to focus on one function of learning rather accommodate the complete set of learning functions. This research will test the proposition that a small set of principles from nonlinear dynamics can produce all of the standard functions of change observed in motor learning. A series of experiments conducted within a dynamical systems framework will examine, in a range of motor tasks, the time scales of change in motor learning. This research will provide a first test of the notion that a unified and parsimonious dynamical account of time scales of change can derive the established set of short- and long-term learning functions in motor learning. The promise of this theoretical and experimental approach to the time scales of motor learning is that it will lead us beyond traditional descriptions of learning toward a predictive science of human motor learning, that links theoretically to neural net approaches to human cognition and artificial system learning.