During rhythmical movement, sensory and motor systems need to interact closely to sustain the rhythm and to meet task requirements. Understanding how our system controls such a basic, all day movement is a prerequisite to improve motor (re)learning models to ameliorate rehabilitation in case of neurological movement disorders, like stroke. Mathematically, the simplest way to model rhythmicity is by means of a continuous oscillator (e.g. Haken et al. 1985). Biological noise interfering with planning and execution makes human movements unavoidably variable, which asks for correction processes (Franklin and Wolpert 2011). One of the principles governing human motor control states that optimised control is characterised by a maximum efficiency, e.g. minimal costs (Guigon et al. 2007). Minimal cost is dependent on the varying interaction between different system characteristics, including anatomical constraints, force generating capacities, and biological noise inducing the intra and interpersonal variability that is inherent to our system’s output (van Galen and Hueygevoort 2000).
Current knowledge about the neural correlates of rhythmical upper limb movement is based on standard finger and wrist movement paradigms that compare different movement conditions within people (high frequency versus low frequency, Kelso et al. 1998; rhythmic versus discrete movements, Schaal et al. 2004). Using this paradigm, simple unilateral rhythmical movements have been shown to elicit contralateral activations of the primary sensorimotor cortex (S1 + M1) and of the supplementary motor area (SMA), complemented by an ipsilateral activation of the anterior cerebellum (Allison et al. 2000; Ball et al. 1999; Schaal et al. 2004). Bilateral movements are associated with a symmetric facilitation of neural activity in the sensorimotor network, with additional frontal activations to ensure coordination between limbs. It is mediated by increased intrahemispheric connectivity and enhanced transcallosal coupling of SMA and M1 (Grefkes et al. 2008; Jäncke et al. 2000).
The activation pattern is comparable between dominant and non-dominant sided movements in extension and intensity when people move at their preferred frequency (Lutz et al. 2005; Jäncke et al. 2000; Koeneke et al. 2004). However, when movement frequency is imposed, activations during non-dominant sided movements increase in intensity compared to those of the dominant side (Lutz et al. 2005). Second, activation increases and expands for both uni and bilateral movements when movement frequency is increased above the preferred frequency (e.g. Kelso et al. 1998; Rao et al. 1996). Together, this demonstrates that moving at a non-preferred frequency is marked by an increase in costs. Therefore, imposing a fixed frequency may lead to different task-induced cost levels between participants and thus lead to biased results when comparing rhythmical motion and its neural correlates between people.
Over the time course of the movement, fine-grained kinematic variables capture the outcome of the interaction between the planned movement and the noise-dependent variability (Newell and Corcos 1993). Here, we explored whether kinematics may additionally provide information on the underlying control system, when the kinematic outcome is linked directly to brain activity. We simultaneously recorded brain activation (fMRI) and movement kinematics during a sensorimotor task that consisted of a self-paced continuous flexion/extension of the elbow. We focused on uni as well as bilateral movements, as many daily living tasks involve bilateral coordination. The task is evaluated as a simple well-known movement that does not require complex motor learning.
Based on the described theoretical model of motor control, we hypothesised that rhythmic voluntary flexion of the elbow is modulated by neural networks involved in (1) the sustained execution of the basic oscillatory rhythmical component and (2) correction processes in reaction to the variability resulting from biological noise. Sustaining the movement in rhythmical motion has been shown to involve the primary sensorimotor network, whereas discrete movements solicit additional higher cortical planning areas (Schaal et al. 2004). First, we expected to confirm the role of the sensorimotor network by performing a standard general linear-model analysis. Second, because task costs were as much equalised over participants as possible, we expected that correlating natural variation in movement execution with variation in BOLD-activation might unmask different brain regions involved in the secondary correction processes that could be (partly) separated from the primary sensorimotor network. […]