Gaze and Gait during Smooth Pursuit While Walking in Older Adults

This dataset contains trial-level data on gaze behaviour, gait performance, and sensory and cognitive functioning in 70 community-dwelling older adults (49 women, 21 men) of which 34 people with fall history in the past 12 months and 36 without. Data were collected as part of a cross-sectional study investigating gaze control during smooth pursuit walking and its relationship with sensory and cognitive function. Data collection was conducted between May 2024 and September 2025 at the Multidisciplinary Motor Centre Antwerp, University of Antwerp, Belgium Study protocol Participants performed a smooth pursuit walking task while wearing an augmented reality headset (Microsoft HoloLens 2). During each trial, participants walked along a 10-metre overground walkway at a self-selected speed while visually tracking a holographic target (hummingbird). The target moved 45° in one of four directions (upward, downward, leftward, or rightward). Each trial comprised an initial stationary target presentation (0.75 seconds), a target movement phase (2 seconds), and a final stationary phase (1 second). Participants completed at least three trials per movement direction, with the order of directions randomised. For more information on the validated procotol (please see: DOI: 10.1016/j.gaitpost.2025.08.001). Gait data were collected using a Vicon T10 motion capture system operating at 100 Hz. A standardized 3D-motion capture protocol including walking with head movements was performed. The marker set consisted of 4 reflective markers onto the augmented reality headset, fixed on the participant's head, as well as 9 reflective markers placed on the following locations: the seventh cervical vertebra, an bilaterally the major trochanders, second metatarsal heads, the calcanea and the lateral malleoli. A familiarization with the setup in standing took place before starting the walking trials. Dataset contents The dataset comprises 950 trial-level observations across 45 variables, organised into the following domains: 1. Participant characteristics Demographic and clinical characteristics were collected during an at-home screening visit: Identification code: a pseudonymised participant code. The prefix C indicates no reported falls, and F indicates one or more reported falls in the past 12 months. Age, in years. Sex, coded as 0 = female and 1 = male. Number of falls: the number of self-reported falls in the past 12 months. Timed Up and Go (TUG): a test of functional mobility, used to confirm that participants could safely complete the walking protocol. Performance was expressed as completion time (s). Montreal Cognitive Assessment (MoCA): a screening test of global cognitive functioning, scored out of 30. Participants scoring ≤21 were excluded. Polypharmacy: recorded as yes/no Education level: highest completed education, on an ordinal scale 1. Primary school not completed 2. Primary school completed 3. Primary school completed and less than 2 years of further education 4. Primary school completed and less than 4 years of further education 5. Vocational secondary education diploma 6. Technical secondary education diploma 7. General secondary education diploma 8. Bachelor's degree (3 years of higher education/university) 9. Master's degree (more than 3 years of higher education/university) 10. Doctoral degree (PhD) Outcome Excel column Unit / coding Identification code ID C = no reported falls, F = ≥1 reported fall Target direction Cond up, down, left, right Age age years Sex gender 0 = female, 1 = male Sex (text) gender 2 Female, Male Number of falls number_of_falls count Timed Up and Go timed_up_and_go_tug s MoCA montreal_cognitive_assessment 0–30 Polypharmacy polypharmacy yes/no yes, no Education level education_level 1–10 2. Gaze behaviour Gaze was recorded using the HoloLens 2 augmented reality headset during a smooth pursuit walking task. Eye and head direction vectors were processed in MATLAB, between target motion onset and the last visible frame of the target, to derive the following outcomes per trial: Cross-correlation function (CCF), reflecting the correspondence between gaze and target trajectories over time. Number of gaze reorientations, representing rapid corrective gaze movements during target tracking. Gaze latency, reflecting the time required to initiate a gaze response to target movement (s). Gaze movement smoothness, quantified using a dimensionless jerk-based measure. Gaze gain: gaze velocity divided by target velocity (ratio) Eye-to-head movement ratio (ratio). Outcome Excel column Unit Cross-correlation function (CCF) CCF_gaze 0–1 Number of gaze reorientations N_Gaze reorientations corrected count Gaze latency latency_gaze s Gaze movement smoothness SmoothnessLDLJ dimensionless Gaze gain gain_gaze ratio Eye-to-head movement ratio Eye_Head_ratio ratio 3. Sensory function Sensory functioning was assessed using three clinical tests targeting cervical, vestibular and visual function: Head Repositioning Accuracy (HRA) test: an assessment of cervical sensorimotor control, specifically position sense, using a head-mounted wearable sensor (3D-Cervical Trainer, Sensamove). Seated (with backrest) and blindfolded participants performed head movements in four directions (flexion, extension, left rotation and right rotation), then returned to a memorised neutral position. There were six repetitions per direction. Performance was quantified as joint position error per direction (°)(DOI: 10.1016/j.msksp.2023.102861). Video Head Impulse Test (vHIT): an assessment of the vestibular semicircular canal function. Participants fixated a target at 1.5 m while passive, high-velocity head impulses were applied. Performance was quantified as vestibulo-ocular reflex (VOR) gain for the horizontal and vertical canals (DOI: 10.3389/fneur.2017.00258). Static visual acuity: participants were seated 1.5 m from a screen and identified the orientation of briefly presented Landolt-C optotypes of progressively adjusted size. Performance was expressed as the visual acuity threshold (logMAR) (DOI: 10.3389/fneur.2022.964017).. Outcome Excel column Unit HRA joint position error, flexion Cerv F ° HRA joint position error, extension Cerv E ° HRA joint position error, left rotation Cerv L ° HRA joint position error, right rotation Cerv R ° VOR gain, horizontal canals (vHIT) VOR HOR gain VOR gain, vertical canals (vHIT) VOR VERT gain Static visual acuity STAT VA logMAR 4. Cognitive function Cognitive functioning was assessed using three neuropsychological tasks targeting distinct cognitive domains: Digit Symbol Substitution Test (DSST): A digitised symbol–number matching task assessing information processing speed and selective attention. Performance was quantified using DSST efficiency, expressed as the time required per correct symbol–number match (ms) (DOI: 10.1097/JCP.0000000000000941). Trail Making Test (TMT): A paper-and-pencil assessment of processing speed, visual scanning, cognitive flexibility, and task-switching. Executive functioning was operationalised using the difference in completion time between Part B and Part A (TMT B–A), expressed in seconds (DOI: 10.1038/nprot.2006.390). Go/No-Go (GNG) task: A digitised response inhibition task assessing sustained attention, inhibitory control, and impulsivity. Participants responded to target stimuli (Go trials) while withholding responses to non-target stimuli (No-Go trials), under progressively shorter response windows. Outcomes included GNG efficiency (time per correct Go response) and the number of commission errors (incorrect responses to No-Go stimuli) (DOI: 10.1037/a0036295). Additionally, global cognitive functioning was assessed using the Montreal Cognitive Assessment (MoCA) as part of participant screening. Outcome Excel column Unit DSST efficiency, levels 1–2 DSST efficiency L1-2 (time needed for 1 correct answer) ms DSST efficiency, levels 1–3 DSST efficiency L1-2-3 (time needed for 1 correct answer) ms TMT B–A TMT B - A s GNG efficiency GNG efficiency P s GNG commission errors not in dataset — MoCA montreal_cognitive_assessment (see section 1) 0–30 5. Gait performance and stability The dataset additionally includes spatiotemporal gait parameters and measures of dynamic stability under tracking and non-tracking conditions. Non-tracking condition: the steps before the onset of the target movement. Outcomes are expressed as the mean across these steps, and variability as the standard deviation (SD) across these steps (named: non-perturbed). Tracking (smooth pursuit) condition: the first step after the onset of the tracking head movement. As this is a single step, no variability measure is available. Later steps were not included due to walkway length limitations (named: perturbed). For both conditions, the following outcomes were calculated: Gait speed: expressed in m/s. Step length: the anteroposterior distance between consecutive heel strikes of opposite feet, expressed in m. Step time: the time between consecutive heel strikes of opposite feet, expressed in s. Step width: the mediolateral distance between consecutive heel strikes of opposite feet, expressed in m. Margins of stability (MoS): the distance between the extrapolated centre of mass and the boundary of the base of support [Hof et al., 2005]. Expressed in m for the backward (BW) and mediolateral (ML) directions; negative values indicate that the extrapolated centre of mass lies outside the base of support. tcome Non-tracking: mean Non-tracking: SD Tracking (first step) Unit Gait speed Mean(GaitSpeed_non perturbed) Std Dev(GaitSpeed_non perturbed) GaitSpeed_perturbed m/s Step length Mean(Steplength_non perturbed) Std Dev(Steplength_non perturbed) Steplength_perturbed m Step time Mean(Steptime_non perturbed) Std Dev(Steptime_non perturbed) Steptime_perturbed s Step width Mean(Stepwidth_non perturbed) Std Dev(Stepwidth_non perturbed) Stepwidth_perturbed m Backward MoS Mean(BW_MoS_non perturbed) Std Dev(BW_MoS_non perturbed) BW_MoS_perturbed m Mediolateral MoS ML_MoS_non perturbed* Std Dev(ML_MoS_non perturbed) Mean(ML_MoS_perturbed)* m Data structure Data are provided in Microsoft Excel format (.xlsx). Each row represents a single smooth pursuit walking trial, identified by participant code and target movement direction. Participant-level demographic, sensory, and cognitive measurements are repeated across the corresponding trial-level observations. The dataset contains observations for all four target movement directions (up, down, left, and right). Missing values are present for some measurements. One sheet (“Dataset gaze-gait smooth pursui…”), 950 rows × 45 columns, in long format. IDs: C1–C37 and F1–F35. C34 and F17 are absent. Cond: the direction of the pursuit target. Right 251, down 235, left 234, up 230. The split is about the same in both groups. Trials per participant: median 13.5 (range 5–30). Only C1 has 30 trials; everyone else has 19 or fewer. Not every participant has all four directions equally often. Reuse potential This dataset may be used to investigate associations between gaze control, sensory and cognitive function, and gait performance in older adults. The repeated-measures structure enables the examination of within-participant and between-participant variability, direction-dependent gaze responses, and potential interactions between gaze behaviour and locomotor stability. The data may also support secondary analyses of gaze–gait coordination, cognitive–motor interactions, and sensory contributions to locomotor control.