The specificity of memory for a highly trained finger movement sequence: Change the ending, change all

Abstract

How are highly trained movement sequences represented in long-term memory? Here we show that the gains attained in the performance of a well-trained sequence of finger movements can be expressed only when the order of the movements is exactly as practiced. Ten young adults were trained to perform a given 5-element sequence of finger-to-thumb opposition movements with their left hand. Movements were analyzed using video based tracking. Three weeks of training resulted, along with improved accuracy, in robustly shortened movement times as well as shorter finger-to-thumb touch times. However, there was little transfer of these gains in speed to the execution of the same component movements arranged in a new order. Moreover, even when the only change was the omission of the one before final movement of the trained sequence (Omit sequence), the initial movements of the sequence were significantly slowed down, although these movements were identical to the initial movements of the trained sequence. Our results support the notion that a well-trained sequence of finger movements can be represented, in the adult motor system, as a singular, co-articulated, unit of movement, in which even the initial component movements are contingent on the subsequent, anticipated, ones. Because of co-articulation related anticipatory effects, gains in fluency and accuracy acquired in training on a specific movement sequence cannot be expressed in full in the execution of the trained component movements or of a full segment of the trained sequence, if followed by a different ending segment.

Introduction

A variety of motor tasks can be conceptualized as consisting of a sequence of simple movement components; the skilled generation of such a sequence would be reduced to the problem of choosing the correct components in the required order, determining the time at which each component movement is initiated, and ensuring smooth continuity from one component movement to the next (Rozenbaum et al., 1983, Jerde et al., 2003). However, there are indications that this scheme may not hold true for highly trained movement sequences. There is evidence supporting the notion that when a certain amount of practice has been afforded in the performance of a given movement sequence, component movements can be concatenated and reorganized into sub-sequences (‘chunked’). Subsequently, in some cases, the sequence may come to be represented as a coherent whole rather than as a set of component movements (Miller, 1956, Verwey, 1994, Karni et al., 1995, Karni et al., 1998, Engel et al., 1997, Blackburn and Young, 2000, Jerde et al., 2003, Sosnik et al., 2004). Extensive practice may, therefore, lead to a representation of movement elements in a qualitatively different manner when part of a given sequence. Thus, in a well-rehearsed motor sequence, movement elements can be influenced by the subsequent, anticipated, movements of the sequence, resulting in spatial and temporal overlap of the movement units. This can generate a new movement entity that is different from the sum of the movement elements that comprised it (Engel et al., 1997, Sosnik et al., 2004). In speech, articulator movements for a given speech sound were shown to vary systematically depending on the nature of the subsequent speech elements and their associated articulator movements (co-articulation, anticipatory effect); a mechanism that supports increasing fluency (Kent and Minifie, 1977, MacNeilage, 1980, Hardcastle and Marchal, 1990, Matthies et al., 2001). Whether fluency in digit movements results in strong anticipatory effects, and whether highly trained finger movement sequences are discreetly represented, in motor memory, as specific sequences, is not clear (Rumelhart and Norman, 1982, Engel et al., 1997, Soechting and Flanders, 1997, Hermsdörfer et al., 2000, Jerde et al., 2003).

Given that well-trained motor sequences can be represented and executed as specific, coherent, movement routines (Karni et al., 1998, Sosnik et al., 2004), with little recourse to feedback and executive control mechanisms (Anderson, 1981, Karni et al., 1998, Chein and Schneider, 2005), the working hypothesis of the current study was that after extensive training the resulting performance gains can be fully expressed only for the sequence as a whole. It is often implicitly assumed that an advantage inherent in the acquisition of skills is that the gains in performance that emerge after practice on a specific task would transfer to the performance of related tasks. However, if extensive training on a movement sequence results in gains in fluency as well as accuracy that reflect strong anticipatory effects, even some small variation on a trained sequence may be executed as a novel sequence. We tested, therefore, the hypothesis that because of the co-articulation-related anticipatory effects, the gains acquired in training on a given sequence would not be expressed in full in the execution of the trained component movements when in a different sequence, or even in the execution of the initial segment (chunk) of the trained sequence if it is to be followed by a different ending segment.