Intentional Constraints on the Dynamics of Human Performance and Behavioral Variability in Motor Control

Intentional Constraints on the Dynamics of Human Performance and Behavioral Variability in Motor Control Auriel Washburn (washbual@mail.uc.edu) Department of Psychology, 4150 Edwards 1 Cincinnati, OH 45221 USA Charles A. Coey (coeyca@mail.uc.edu) Department of Psychology, 4150 Edwards 1 Cincinnati, OH 45221 USA Michael J. Richardson (richamo@mail.uc.edu) Department of Psychology, 4150B Edwards 1 Cincinnati, OH 45221 USA Abstract Manipulation of environmental constraints has been shown to influence the relative amounts of voluntary and involuntary control employed by a person to complete a task, as well as the resulting structure of performance variability. Generally, the voluntary control required when no constraints are present leads to self-similar changes in performance, some constraint provides involuntary control that leads to random fluctuations in performance, and constraint which provides feedback about performance accuracy can result in anti-persistent variability. The current study investigated whether providing two groups of individuals with different intentions for the same task would produce changes in voluntary and involuntary control similar to that observed following the manipulation of task constraints. Results indicated that a difference in intention does result in divergent uses of voluntary and involuntary control and distinctly different structures in performance variability. Key words: intention; fractal structure; voluntary and involuntary control; motor control Over the past decade, a substantial amount of research has focused on determining what information can be gained about human cognitive and motor processes by assuming that they are inextricably linked through what is often referred to as the ‘interaction-dominant dynamics’ of human behavior (Holden, Van Orden, & Turvey, 2009; Ingber, 2003; Turvey, & Moreno, 2006; Van Orden, & Holden, 2002; Van Orden, Holden, & Turvey, 2003; Van Orden, Holden, & Turvey, 2005). As noted by Van Orden (2010), absolute independence of these processes would allow for random variability in performance within each process, while dominance by one process over all others would cause highly regular fluctuations across processes. Standard, linear statistical methods for assessing performance are based on an assumption of random variability, or noise, in performance and, necessarily, the belief that whatever process is being evaluated can be thought of as independent from all other contemporaneous processes. However, methods for assessing potential structure within variability over time reveal that while fluctuation in performance is sometimes strictly random, more often variability is characterized by patterns occurring at a variety of different timescales (Ferrer-i-Cancho & Elvevag, 2010; Kiefer, Riley, Shockley, Villard, & Van Orden, 2009; Eke, Herman, Kocsis, & Kozak, 2002; Eke, Herman, Bassingthwaighte, Raymond, Percival, Cannon, Balla, Ikrenyi, 2000; Gilden, 2001; Holden et al., 2009; Kuznetsov & Wallot, 2011; Phillipe, 2000; Rhodes & Turvey, 2007; Wallot & Van Orden, 2011a, b; Warren, Carciun, & Anderson--Butcher, 2005). This type of variability is neither strictly random, nor strictly regular, but is rather somewhere in between the two, and therefore suggestive of both competitive and cooperative interactions between the different cognitive and motor aspects of the behavior under observation (Van Orden, 2010). The patterned variability in performance described above is defined by a fractal structure, in that self-similarity in fluctuations is apparent at multiple timescales (Mandelbrot, 1982; Brown & Liebovitch, 2010; West & Deering, 1995). This type of variability is typically referred to as ‘pink’ noise, in contrast to the ‘white’ noise of random fluctuation (Van Orden, 2010). In order to determine what kind of variability is occurring for a given task, it is important to repeatedly measure some aspect of that task as performance unfolds over time. The resulting series can then be broken down into several composite, sinusoidal series each with a different amplitude and frequency. A Power-Spectral Density (PSD) analysis can then be used to give an assessment of variability (Delignieres, Ramdani, Lemoine, Torre, Fortes, & Ninot, 2006; Holden, 2005; Marmelat & Delignieres, 2011). The slope of a regression line fit to a plot of the logarithm of the power (amplitude squared) of changes with the logarithm of their corresponding frequencies provides a unique scaling relation between the size and frequency of changes in the performance time series. This scaling relation (S) is related to a characteristic scaling exponent (α), where α = -S (Holden, 2005). It is this scaling exponent which is used to give a qualitative assessment of the type of variability being observed. Since there will be no systematic relationship between the size and