# Mapping the complexity of motor variability: From individual space of variability to motor fingerprints

## TL;DR

This paper introduces a new geometric method to study motor variability, revealing unique individual motor signatures based on how people vary their movements.

## Contribution

A novel geometric approach to model motor variability using Procrustes transformation and multidimensional scaling.

## Key findings

- The method defines an individual space of variability based on bidimensional data dispersion.
- It identifies unique motor fingerprints by analyzing how individuals vary their kinematic parameters.
- The approach generalizes across different datasets and sensor types.

## Abstract

Traditional average-based metrics have long been considered the gold standard in behavioral and brain research. However, recent advances emphasize the importance of examining the dispersion around the mean to uncover the nuances of individual differences and challenge simplistic assumptions. Thus, the study of variability is becoming increasingly central across a wide range of domains. Here, we tackle the composite architecture of motor variability by proposing a new geometric method to model it. Three independent gait datasets are used to: i) develop the method (Dataset 1), ii) evaluate its performance when transitioning from optoelectronic cameras to inertial measurement units (Dataset 2), and iii) generalize it in an experimental design with cognitive manipulations (Dataset 3). The method is based on the Procrustes transformation and multidimensional scaling. This geometric approach allows us to define the individual space of variability (i.e., the amount of bidimensional space covered by each individual’s trial-by-trial data). In turn, it provides robust evidence to identify each individual unique and specific motor signature (motor fingerprint). Our approach represents a fundamental shift from previous research: It is not the value of kinematic parameters per se that defines an individual’s motor signature, but rather the distinct way in which each individual varies these parameters, i.e., the dispersion of the distribution of their kinematic data. This novel perspective provides a single-subject-weighted proxy for motor signature, based on the characteristic dispersion of each individual’s data. The potential applications of this new method in research and clinical settings represent a fascinating future challenge.

The online version contains supplementary material available at 10.3758/s13428-025-02635-0.