Dual Adaptation Supports a Parallel Architecture of Motor Memory

Abstract

Although our understanding of the mechanisms underlying motor adaptation has greatly benefited from previous computational models, the architecture of motor memory is still uncertain. On one hand, two-state models that contain both a fast-learning-fast-forgetting process and a slow-learning-slow-forgetting process explain a wide range of data on motor adaptation, but cannot differentiate whether the fast and slow processes are arranged serially or in parallel and cannot account for learning multiple tasks simultaneously. On the other hand, multiple parallel-state models learn multiple tasks simultaneously but cannot account for a number of motor adaptation data. Here, we investigated the architecture of human motor memory by systematically testing possible architectures via a combination of simulations and a dual visuomotor adaptation experimental paradigm. We found that only one parsimonious model can account for both previous motor adaptation data and our dual-task adaptation data: a fast process that contains a single state is arranged in parallel with a slow process that contains multiple states switched via contextual cues. Our result suggests that during motor adaptation, fast and slow processes are updated simultaneously from the same motor learning errors.

Savings in locomotor adaptation explained by changes in learning parameters following initial adaptation

“…Different variations of the SSM have been recently suggested to explain adaptation and savings during force field (Donchin et al 2003;Smith et al 2006), object rotation (Ingram et al 2011), and visuomotor (Lee and Schweighofer 2009;Zarahn et al 2008) perturbations. Most of these models assume linear time invariant (LTI) properties of the parameters (Donchin et al 2003;Ingram et al 2011;Lee and Schweighofer 2009;Smith et al 2006) while the rest model assumes varying parameters that change with experience (Berniker and Kording 2011;Zarahn et al 2008). All of these error-based models suggest that trial-by-trial adaptation occurs by updating the appropriate internal models (i.e., states) to reflect the behavior of the perturbation.…”

“…Initial attempts to model adaptation to an external perturbation were based on state space models (SSMs) composed of a fast and one or multiple slow processes (Lee and Schweighofer 2009;Smith et al 2006). However, these linear multiple-rate SSMs could not explain savings that occur after a prolonged period of washout (Krakauer et al 2005;Zarahn et al 2008) and across days (Robinson et al 2006).…”

“…However, these linear multiple-rate SSMs could not explain savings that occur after a prolonged period of washout (Krakauer et al 2005;Zarahn et al 2008) and across days (Robinson et al 2006). Instead, a recently nonlinear SSM (Zarahn et al 2008) and context-dependent models (Ingram et al 2011;Lee and Schweighofer 2009) were suggested to better explain a variety of phenomena reported in the motor adaptation literature, including savings. While evidence for savings has been accumulated from different systems [saccades, arm reaching, and locomotion (Kojima et al 2004;Krakauer et al 2005;Malone et al 2011)] and across paradigms [saccades, visuomotor; and force filed adaptation (Kojima et al 2004;Krakauer et al 2005;Smith et al 2006;Zarahn et al 2008)], adaptation and savings were mainly modeled based on reaching and saccades adaptation results and, to the best of our knowledge, were never modeled for locomotor adaptation.…”

Gain Field Encoding of the Kinematics of Both Arms in the Internal Model Enables Flexible Bimanual Action

“…We modeled the motor learning process using the state space model (Thoroughman and Shadmehr, 2000;Wainscott et al, 2005;Lee and Schweighofer, 2009;Nozaki and Scott, 2009), consisting of N primitives that encode the movement directions of both arms as g( r , l ), where r and l are the movement directions of the right and left arms, respectively (see Fig. 3A).…”

Visual error augmentation enhances learning in three dimensions

“…Other investigators have found cases in which learning could not be described in terms of exponential decay functions, where "no meaningful value for τ could be calculated" and "the problem could not be alleviated by using double exponentials [6]." However, others suggest two [7] (Smith and Shadmehr) or more [8] (Schweighofer) learning processes. Our data suggests immediate performance changes in the error-augmented group when compared to the control group.…”