Temporal structure in sensorimotor variability: a stable trait, but what for?

Abstract

Human performance shows substantial endogenous variability over time. Such variability has been known to show temporal structures: Performance from action to action is not independent, but shows correlation with itself over time. While the existence of such dependencies has been frequently reported on, its measurement and interpretation come with a number of controversies, and its potential benefit for studying individual differences remains unclear. Two recent studies have linked temporal structures to individual differences in task performance, but with contrasting results. In the current study, we aim to investigate the intra-individual repeatability of these temporal structures in endogenous performance on the Metronome Task (25 participants, tested in two sessions ~45 minutes apart). Secondly, we examine the inter-individual correlates of the temporal structures (83 participants), specifically looking at: 1) task performance, 2) meta-cognitive ratings of attentional state, and 3) self-assessed personality traits (ADHD tendencies, mind wandering, and impulsivity).Rather than using one analysis method (as is common in the literature), we consistently compare all the frequently-used analysis methodsallowing us to investigate the structures without an a prior assumption on the underlying time scales.Results indicate that autocorrelation at lag 1 and Power Spectra Density slope showed the most intra-individual repeatability, ARFIMA(1,d,1) parameters showed the least, and evidence for Detrended Fluctuation Analysis was indeterminate with a moderate effect size at best. Overall, the autocorrelation at lag 1 seemed the best measure for studying individual differences, due to its high reliability and ease of use. Furthermore, with exception of the ARFIMA parameters, temporal structure was correlated with performance but not with subjective attentional state or self-assessed personality traits between participants.

Variance (un)explained: Experimental conditions and temporal dependencies explain similarly small proportions of reaction time variability in linear models of perceptual and cognitive tasks

“…It is possible that temporal predictability is higher in sensorimotor actions other than the ones we scrutinized in the present study, particularly in tasks without experimental conditions such as tapping to a metronome (see Perquin et al, 2022 for an overview of autocorrelations in RT series from various tasks with and without experimental conditions). However, even in these tasks without experimental conditions, the largest proportion of the RT variance remained unexplained – meaning the trial-to-trial responses were still largely unpredictable.…”

“…Both behavioural variability such as the variance across an entire session (Hultsch et al, 2002; Hultsch et al, 2000; Perquin & Bompas, 2019; Perquin et al, 2022; Saville et al, 2011; 2012) as well as temporal dependencies in entire RT series (Perquin et al, 2022) exhibit considerable stability across time and are thus trait-like. This makes them well-suited for research about individual differences.…”

“…However, temporal dependencies are also present in tasks in which the same sensorimotor action is performed repeatedly and without any changes in experimental condition, such as tapping to a fixed interval (e.g., Chen et al, 2002;Lemoine et al, 2006;Madison, 2001;Perquin et al, 2022). Because the external environment remains the same, such dependencies must be driven endogenouslyi.e., arising from fluctuations in the internal system.…”

“…Another line of research emphasises that the trial-by-trial variability of behaviour is not random, but exhibits structured dependencies over time (e.g., Gilden, 2001; Van Orden et al, 2003; Perquin et al, 2022; Torre et al, 2019; Torre & Wagenmakers, 2009; Wagenmakers et al, 2004; Welford, 1968). For example, most RT series show positive autocorrelations (Figure 1E), indicating that fast RTs are more likely followed by fast than slow RTs, and vice versa.…”

“…An example is the ‘sequential congruency’ (alternatively, ‘Gratton’) effect, which denotes the phenomenon that participants respond faster if the previous trial was of the same condition as the current one (e.g., both congruent in the Flanker task; Figure 1F). However, temporal dependencies are also present in tasks in which the same sensorimotor action is performed repeatedly and without any changes in experimental condition, such as tapping to a fixed interval (e.g., Chen et al, 2002; Lemoine et al, 2006; Madison, 2001; Perquin et al, 2022). Because the external environment remains the same, such dependencies must be driven endogenously – i.e., arising from fluctuations in the internal system.…”