Examining the Influence of Task Complexity on Dual-Task Performance in Parkinson’s Disease and Healthy Older Adults

Abstract

Introduction

People with Parkinson’s disease (PwPD) often experience difficulty performing dual-tasks (DTs), negatively impacting mobility, fall risk, and quality of life. The influence of task complexity on DT performance remains unclear. This study examined the effects of gait complexity and cognitive/speaking task on DT effects (DTEs) in PwPD and healthy older adults.

Methods

Forty-one PwPD and eleven healthy older adults completed two gait tasks – straight walking (simple) and figure-8 walking with obstacle crossing (complex) – on a pressure-sensitive walkway, paired with four cognitive/speaking tasks: oral trail making, counting, diadochokinetic task, and spontaneous speech. Each combination was performed three times for 20 s. Single-task and DT performance were assessed for gait parameters (step length, step length coefficient of variation) and cognitive/speaking task response rate. Motor DTE (mDTE), cognitive DTE (cogDTE), combined DTE (cDTE), and modified attention allocation index (mAAI) were calculated. A 2 × 5 repeated-measures ANOVA tested effects of gait complexity and cognitive/speaking task on raw gait metrics, and a 2 × 4 ANOVA tested effects on DTE metrics.

Results

In PwPD, complex walking resulted in shorter step length and greater step length coefficient of variation compared to simple walking (ps < 0.001), whereas cognitive/speaking task did not affect raw gait metrics. For DTEs, cognitive/speaking task influenced step length-based mDTE, cDTE, and mAAI (ps < 0.05) and step length coefficient of variation-based mAAI (p < 0.05). In healthy older adults, gait complexity led to greater step length coefficient of variation (p < 0.001) while step length was unchanged. Cognitive/speaking task significantly influenced step length-based cogDTE, cDTE, and mAAI, as well as step length coefficient of variation-based cogDTE and mAAI (ps < 0.05). No significant interactions were observed.

Conclusion

In PwPD, gait complexity influenced raw gait performance, whereas DTEs were more sensitive to cognitive/speaking task. In healthy older adults, gait complexity primarily affected step length variability, while DTEs were also influenced by cognitive/speaking task. These findings emphasize the value of diverse DT paradigms and the importance of assessing both raw performance and DTEs. Tailoring assessments to include a variety of motor and cognitive challenges may improve sensitivity in evaluating gait and fall risk, supporting more personalized interventions.

Introduction

Dual tasking (DT) involves the performance of two tasks concurrently, often one motor and one cognitive [1–3]. This requires distribution of attention similar to that often seen in daily life activities such as walking while talking with a friend. Performing a DT can provide insights into one’s automaticity, or the ability to perform a task without directing attention to its execution [1, 4]. However, even in healthy individuals without apparent motor or cognitive deficits, finite processing capacity [5] can compromise the execution of a DT [4]. This can be observed in DT-related changes in spatiotemporal gait parameters, such as slower gait speed, shorter stride length, and increased stride variability [3, 6]. These changes in motor task performance with the addition of a secondary cognitive task relative to single-task (ST) performance are indicative of cognitive-motor DT interference [1, 6–8].

Individuals with neurodegenerative diseases often experience greater difficulty performing DTs [4, 9]. In particular, people with Parkinson’s disease (PwPD) commonly exhibit automaticity-related motor deficits such as decreased stride length, reduced gait velocity, and increased gait variability when performing DTs compared to healthy older adults [3, 4, 6]. These DT challenges can have real-world consequences as impaired DT performance has been linked to an increased risk of falls, greater disability, and challenges in completing everyday activities [10, 11]. Furthermore, DT performance has been shown to be related to disease severity and functional independence in PwPD [10, 11].

There have been few studies that comprehensively assess how various primary and secondary task demands influence DT performance. Overall, evidence suggests that secondary task type, as well as the specific demands it places on cognition or motor control, can influence DT performance, with some tasks causing greater disruption in gait or other motor outcomes than others [6, 12, 13]. A more recent study also found that walking condition significantly affected walking performance in individuals with Parkinson’s disease further supporting the influence of both task demands on performance [14]. The type of cognitive task appears to influence changes in specific gait parameters, such as arm swing or stride characteristics, although some gait changes, such as reductions in gait velocity and joint range of motion, appear to occur regardless of the secondary task employed [15, 16]. Evidence also suggests that variations in motor task demands can influence overall DT performance. For example, engaging in certain concurrent motor tasks may facilitate or interfere with cognitive performance under DT conditions, although the effects of different walking demands remain underexplored [17]. Importantly, the complexity and novelty of both the primary and secondary tasks affect the level of attention directed toward each task, as evidenced by cortical processing demands varying with task difficulty [18, 19]. This is particularly relevant given that gait, especially DT gait, is a complex behavior drawing on motor, sensory, and cognitive systems [20]. For instance, cognitive tasks have been shown to disrupt gait more than secondary motor tasks [3, 21], and different cognitive domains appear to uniquely influence gait characteristics [3, 18]. In addition, the posture-first hypothesis suggests that individuals tend to prioritize gait or postural stability when attentional resources are limited, thereby protecting bodily integrity during challenging conditions [7, 22]. Healthy older adults commonly demonstrate this motor prioritization, whereas PwPD may show reduced or inconsistent prioritization due to impaired automaticity, making gait more susceptible to interference under DT demands [22, 23]. Together, these findings highlight critical gaps in literature, particularly the lack of direct and comprehensive comparisons across both cognitive domains and motor task complexities. This underscores the need to clarify how specific cognitive domains interact with various movement demands under DT conditions, especially in neurologically compromised populations such as PwPD.

To this end, the purpose of this study was to explore differences in DT abilities across various task types and levels of complexity in PwPD. Specifically, we aimed to determine whether gait complexity, defined as simple (straight-line walking) versus complex (figure-eight walking with an obstacle), influenced step length, step length variability, and dual-task effects (DTEs) during DT conditions. We also investigated how different cognitive/speaking tasks with distinct underlying cognitive demands affected these outcomes. Additionally, we assessed whether there were interaction effects between gait complexity and cognitive/speaking task on gait and DTEs in both groups. A secondary purpose of this study was to examine differences in DTEs between PwPD and healthy older adults. We hypothesized that both PwPD and healthy older adults would exhibit shorter step length and greater step length variability during complex gait paradigms compared to simple straight-line walking, as well as greater motor and cognitive DTEs, indicating increased interference on both the primary gait task and the secondary cognitive/speaking task. We further hypothesized that different cognitive/speaking tasks would differentially impact gait performance and DTEs, with both groups showing task-specific variations in step length, step length variability, and DT performance with greater DT costs in PwPD.

Methods

Participants

Forty-one PwPD and eleven healthy older adults participated in this study. Sample sizes differed because the study was primarily designed to examine DT performance in PwPD, with healthy older adults included for contextual comparison. All participants provided informed consent as approved by the Institutional Review Board at Saint Louis University. PwPD were under the care of movement disorder specialists. PwPD who were taking dopaminergic medication (n = 38) were tested in their on-medication state, defined as the period when dopaminergic medication was active and providing optimal motor symptom relief. If a participant reported not being in the on-medication state at the start of testing, data collection was paused until the on state was achieved. A small subset of participants with Parkinson’s disease (n = 3) were not taking dopaminergic medication due to early disease stage or clinical management decisions. All participants were native speakers of American English and had adequate hearing and vision, with correction (e.g., glasses or hearing aids) as needed. Exclusion criteria for PwPD included a diagnosis of another neurological disorder, inability to ambulate at least 9 meters independently without the use of assistive devices (e.g., canes or walkers), or significant cognitive impairment (Mini-Mental State Exam score ≤20). Healthy older adults reported no history of neurological, cognitive, speech, or movement disorders.

Design

This study employed a cross-sectional design to examine DT abilities in both PwPD and healthy older adults. All testing was completed during a single laboratory visit. Screening procedures, balance ability testing, and disease severity assessments were conducted prior to a DT paradigm, which systematically varied both gait complexity and cognitive/speaking task demands. The DT paradigm and associated outcome measures are detailed below.

Instrumentation

Disease Severity

Disease severity in PwPD was assessed using the Movement Disorder Society-Unified Parkinson’s Disease Rating Scale, Part III (MDS-UPDRS III), which evaluates motor symptoms including rigidity, tremor, bradykinesia, and gait/postural instability. The MDS-UPDRS has been validated as a reliable and internally consistent measure with strong correlations to the original UPDRS, and its Part III motor examination is supported for use as an index of Parkinson’s disease severity [24]. Higher scores indicate greater Parkinson’s disease-related motor impairment. In addition, disease stage was classified using the Hoehn and Yahr scale, a widely used clinical rating that provides a global estimate of Parkinson’s disease severity based on motor impairment and disability, and which, despite its relative simplicity, has demonstrated strong clinometric performance and ease of application, supporting its long historical acceptance as a concise summary of disease status [25].

Balance Ability

Balance performance was assessed using the Mini-Balance Evaluation Systems Test (Mini-BESTest), a standardized clinical measure of dynamic balance that evaluates anticipatory postural adjustments, reactive postural control, sensory orientation, and dynamic gait [26]. The Mini-BESTest was administered as part of the standardized data collection protocol to characterize balance ability in participants. Mini-BESTest scores were not used as inclusion or exclusion criteria.

DT Paradigm

To examine how motor and cognitive demands interact during walking, participants completed a DT paradigm in which gait was paired with a variety of cognitive/speaking tasks. This design enabled systematic evaluation of how increasing gait complexity and different cognitive task demands influenced DT performance. Participants completed two gait tasks of differing complexity, illustrated in Figure 1. The simple gait task consisted of straight-line walking for 13 meters with turns at each end. The complex gait task required participants to walk a figure-eight path around two cones placed 4.9 meters apart while stepping over a 15.2-centimeter obstacle positioned at the center of the Zeno™ Walkway during each pass. The complex gait task was designed to impose qualitatively different motor demands by combining continuous walking, obstacle negotiation, and planned directional changes, rather than to isolate the effects of any single component.

Fig. 1.

Design of the DT paradigm with simple and complex gait tasks. Participants completed two gait tasks of differing complexity while performing cognitive/speaking tasks. The simple gait task involved straight-line walking for 13 m with turns at each end. The complex gait task required walking a figure-eight path around two cones placed 4.9 m apart, stepping over a 15.2-cm obstacle at the midpoint of each pass.

During both gait tasks, participants performed four cognitive/speaking tasks designed to engage distinct cognitive or speech motor processes. The first task was the Oral Trail Making Test (oTMT), which primarily assesses executive function and set-shifting [27]. Participants verbally alternated between letters and numbers in ascending order (e.g., B-3, C-4, D-5), with the starting letter/number pair varied across blocks to reduce familiarity effect. Each oTMT trial lasted 20 s per trial for 3 trials. The second task was an automatic counting task intended to represent low cognitive load. Participants counted aloud from 1 to 20 at their normal pace and restarted at 1 after each sequence until the end of the trial. Each automatic counting trial lasted 20 s per trial for 3 trials. The third task was the diadochokinetic (DDK) task assessing articulatory speech and coordination through rapid syllable repetition [28, 29]. Participants were instructed to repeat single syllables (/pa/, /ta/, /ka/) as well as a sequence of the three syllables (/pataka/) as quickly and as clearly as possible. Each DDK trial lasted 40 s and consisted of four consecutive 10-s segments: /pa/(10 s), /ta/(10 s), /ka/(10 s), and /pataka/(10 s). Three DDK trials were completed. The fourth task was a spontaneous speech task requiring natural lexical retrieval and narrative fluency [30, 31]. Participants spoke continuously about a familiar topic of their choice (e.g., hobbies, family, sports, or pets) for 20 s per trial for 3 trials. Verbal responses were recorded for all cognitive/speaking tasks for subsequent analyses. These tasks were selected to represent distinct cognitive processes rather than to systematically manipulate overall task difficulty.

Participants first completed cognitive/speaking tasks as STs while standing to establish baseline cognitive/speaking task performance. Cognitive/speaking tasks were administered in a fixed order (no cognitive/speaking task, oTMT, automatic counting, DDK, and spontaneous speech), with additional periods of no cognitive/speaking task (“rest”) interspersed between each task. This sequence was completed in three consecutive blocks to obtain multiple measurements of each task condition and improve the stability of the resulting gait and cognitive/speaking task performance estimates. The same procedure and sequence of cognitive/speaking tasks was then repeated while performing the simple and complex gait tasks, during which participants walked continuously without stopping. ST walking performance was derived from the “no cognitive/speaking task” periods during both gait conditions, in which participants walked without speaking. Repeating the full sequence across both simple and complex gait tasks ensured that gait and cognitive/speaking task performance could be evaluated consistently across different levels of motor demand. Seated rest periods were provided between ST, simple gait, and complex gait conditions, and participants resumed testing once they reported feeling sufficiently recovered to continue. During DT conditions, participants were instructed to perform both the walking and cognitive/speaking tasks to the best of their ability, and no explicit instructions regarding task prioritization were provided.

Gait and Cognitive Measures

Spatiotemporal gait parameters were collected using the Zeno™ Walkway Gait Analysis System (ProtoKinetics LLC, Havertown, PA, USA), a 6-meter by 1.2-meter sensor-embedded mat that records pressure data. The device and its software have demonstrated strong test-retest reliability and strong concurrent validity with the GAITRite walkway, a widely validated and extensively used tool in gait research [32, 33]. For the simple gait task, only straight-line walking was analyzed, with turns completed off the walkway. For the complex gait task, turning segments were also excluded so that only continuous walking was evaluated. If a freezing of gait episode occurred during a pass, that entire pass was removed from analysis. However, freezing was observed only during turning segments during complex gait conditions, which were excluded a priori from all analyses. As a result, no additional passes were excluded due to freezing. Primary gait outcomes included step length and step length variability. Step length variability was quantified using the coefficient of variation, calculated as the standard deviation of step length divided by the mean step length and multiplied by 100. Step length and step length variability (step length coefficient of variation) were selected because both have been previously linked to impaired motor automaticity and DT-related gait deficits in PwPD [3]. Together, these measures provide complementary information: step length reflects overall gait magnitude, while step length coefficient of variation reflects the quality and regularity of walking.

Cognitive performance during each cognitive/speaking task was quantified using response rate, defined as the number of correct responses per second (for oTMT) or production rate (for automatic counting, DDK, and spontaneous speech). Response rate was selected because it provides an objective, quantifiable measure of task performance, allowing DTEs to be calculated.

DTE Battery

DTE metrics, including motor DTE (mDTE), cognitive DTE (cogDTE), combined DTE (cDTE), and modified Attention Allocation Index (mAAI), were computed separately for two gait features: step length and step length coefficient of variation. These metrics were calculated across four cognitive/speaking tasks (oTMT, automatic counting, DDK, and spontaneous speech) and two walking complexities (simple and complex), yielding a total of eight DTE measures per gait feature, or sixteen per participant. The calculations were based on previously established approaches [1].

mDTE: mDTE quantifies the relative change in motor performance under DT conditions:

$$mDTE\left(\%\right)=\frac{motorDT-motorST}{motorST}\times 100\%$$

For step length, longer steps reflect better performance, so values were multiplied by + 100%. For step length coefficient of variation, higher values indicate more variable performance, generally considered worse performance, consequently values were multiplied by −100% to maintain consistent interpretation. Positive mDTE values indicate facilitation of motor performance under DT conditions, whereas negative values indicate interference.

cogDTE: cogDTE quantifies the relative change in cognitive performance under DT conditions:

$$cogDTE\left(\%\right)=\frac{cognitiveDT-cognitiveST}{cognitiveST}\times 100\%$$

Higher task rates indicate better cognitive performance, so cognitive measures were multiplied by + 100%. Positive cogDTE values indicate facilitation of cognitive performance under DT compared to ST conditions, while negative values indicate a decline in cognitive performance during the DT condition.

cDTE: cDTE represents total DT interference by integrating motor and cognitive performance. This measure allows for a more comprehensive assessment of DT performance than mDTE or cogDTE alone as it integrates performance across both domains. By combining motor and cognitive outcomes, cDTE provides a single, interpretable metric of overall DT ability. It was calculated using the following formula:

$$cDTE\left(\%\right)=\frac{\left(motorDT\times cognitiveDT\right)-\left(motorST\times cognitiveST\right)}{\left(motorST\times cognitiveST\right)}\times 100\%$$

This metric captures the overall effect of performing two tasks simultaneously. For cDTE calculations using step length as the motor performance metric, both motor and cognitive measures are in the “higher = better” direction, so no inversion was required. For cDTE calculations using step length coefficient of variation, where higher variability is generally accepted as “worse” motor performance [34, 35], cognitive measures were inverted (seconds per correct response) to align directionality with motor performance, and the resulting values were multiplied by −100% instead of +100%. This maintains interpretability such that positive cDTE values consistently indicate facilitation, whereas negative values indicate interference.

mAAI: mAAI quantifies how attentional resources are distributed between motor and cognitive tasks:

$$mAAI=mDTE-cogDTE$$

Positive mAAI values indicate a greater relative prioritization of the motor task, whereas negative values indicate greater attention on the cognitive task.

Statistical Analyses

All statistical analyses were conducted using R (version 4.3.2).

Imputation

Missing values occurred across clinical variables (disease duration, Mini-BESTest, MDS-UPDRS III) and outcome variables (step length, step length coefficient of variation, and cognitive/speaking task response rates). This primarily arose from isolated technical issues, including incomplete walkway sensor recordings or missing audio files, as well as occasional documentation errors in clinical assessments, and was not attributable to participant noncompliance or inability to complete the tasks.

Because clinical severity, gait performance, and cognitive/speaking task performance are theoretically and empirically interrelated, all clinical, gait, and behavioral variables were included as predictors in the imputation model. Imputation was performed using predictive mean matching implemented in the mice package in R. To maintain comparability, imputations were conducted separately for PwPD and healthy older adults and separately for simple and complex walking conditions, such that values were estimated using data pooled across participants within each subgroup. Given the low proportion of missingness and its predominantly technical origin, imputation was selected instead of excluding participants with partial missing data to reduce bias and retain statistical power.

Repeated-Measures Analyses of Gait and DTE Outcomes

To examine the effects of walking task complexity (2 levels: simple, complex) and speaking task condition (5 levels: ST walking, oTMT, automatic counting, DDK, spontaneous speech) on raw gait outcomes (step length and step length coefficient of variation), 2 × 5 repeated-measures ANOVAs were conducted. These models allowed evaluation of both the main effects of task complexity and cognitive/speaking task, as well as their interaction, while accounting for the within-subject correlations inherent in repeated measurements.

Separate 2 × 4 repeated-measures ANOVAs were conducted for DTE outcomes (mDTE, cogDTE, cDTE, mAAI), which operationally excluded the ST conditions. Each model included complexity and condition as within-subject factors, and analyses were performed separately for each gait parameter (step length and coefficient of variation).

All ANOVA models were conducted once within the PwPD group and once within the healthy older adult group, treating complexity and cognitive/speaking task as within-subject factors, reflecting that each participant completed every combination of gait complexity and speaking task. Prior to inferential testing, normality of outcome variables was evaluated within each group using Shapiro-Wilk tests; although some variables demonstrated departures from normality, repeated-measures analyses of variance were retained given their established robustness to moderate violations of normality in within-subject designs and to allow consistent interpretation of main effects and interactions across task conditions. Denominator degrees of freedom were adjusted using a Satterthwaite-style approximation to account for repeated measures. Significance was set at p < 0.05, and post hoc comparisons were performed only for significant main effects or interactions to control for type I error. Effect sizes (partial eta-squared, η²p) were also reported to provide a measure of the magnitude of observed main effects. Conventional benchmarks (≈0.01 small, 0.06 medium, 0.14 large) are provided for interpretive context; however, these values should be interpreted cautiously as partial eta-squared is influenced by study design and within-subject variance [36]. For all repeated-measures ANOVAs, post hoc pairwise comparisons were performed only when a significant main effect or interaction was observed. Bonferroni correction was applied to all post hoc comparisons to control for type I error associated with multiple testing. Cohen’s d was calculated for post hoc pairwise comparisons to quantify effect size magnitude and was interpreted using conventional thresholds, with values of approximately 0.2 considered small, 0.5 medium, and 0.8 large [36, 37].

Group Comparisons of DTEs

To compare DTE between Parkinson’s disease and healthy older adults, separate Welch’s ANOVAs were conducted for mDTE, cogDTE, cDTE, and mAAI, with group (PwPD vs. healthy older adults) as the between-subjects factor. Welch’s ANOVA was used to account for unequal variances and unequal sample sizes between groups. This approach was selected a priori given the substantial difference in sample sizes between groups, and Levene’s tests indicated heterogeneity of variances for several DTE outcomes, further supporting the appropriateness of Welch’s method. If a significant group effect was observed, pairwise comparisons were to be performed using Welch’s t tests with Bonferroni correction.

Results

Sample Characteristics

Table 1 presents sample characteristics for the Parkinson’s disease and healthy older adult groups, along with the gait and speech metrics included in this analysis. ST gait step length and step length coefficient of variation for each group are also reported.

Table 1.

Demographic and clinical characteristics of PwPD and HOA

​	HOA (N = 11)	PwPD (N = 41)
Age, years	66.73 (61.64, 71.81)	68.05 (65.66, 70.44)
Sex, n (%)
Male	6 (54.5)	24 (58.5)
Female	5 (45.5)	17 (41.5)
Race, n (%)
White	11 (100)	38 (92.7)
Black or African American	NA	3 (7.3)
Disease duration, months	0.00 (0.00, 0.00)	42.26 (28.07, 56.44)
MDS-UPDRS III	3.73 (0.84, 6.61)	28.59 (24.21, 32.97)
Hoehn and Yahr scale, n (%)
Stage 0	11 (100)	NA
Stage 1	NA	7 (17.1)
Stage 2	NA	32 (78)
Stage 3	​	2 (4.9)
Mini Balance Evaluation Systems Test	25.36 (24.33, 26.40)	22.12 (21.03, 23.22)
Mini-Mental State Examination	28.73 (27.67, 29.79)	27.98 (27.20, 28.75)
Step length (during simple ST Walking), cm	65.26 (60.93, 69.58)	60.78 (58.37, 63.19)
Step length coefficient of variation (during simple ST Walking)	5.58 (3.44, 7.71)	6.16 (4.60, 7.73)
Step length (during complex ST Walking), cm	63.65 (58.51, 68.79)	56.28 (53.95, 58.62)
Step length coefficient of variation (during complex ST Walking)	11.27 (8.91, 13.63)	14.80 (13.00, 16.60)

Continuous variables are reported as means with 95% CI, and categorical variables are presented as counts with corresponding percentages.

Data Completeness

Overall data completeness was high. Overall, 73.1% of participants had complete datasets with no missing values across clinical, gait, and cognitive/speaking task measures. Across the full dataset, 125 values were missing, representing 5.86% of all observations, and no individual variable exceeded 13.5% missingness. At the participant level, the proportion of imputed data ranged from 0% to 30%, with the upper bound reflecting missing recordings for an entire gait condition (simple or complex) due to isolated technical failures rather than scattered missing observations. There were no participant dropouts. Freezing of gait was observed during turning in the complex walking condition in two participants; however, turning segments were excluded a priori from complex gait analyses, and no straight-walking trials were excluded due to freezing. Missing values were addressed using multiple imputation as described in the Methods.

Step Length Analyses

PwPD Group

Step length raw values: a 2 (gait complexity: simple, complex) × 5 (cognitive/speaking task: ST walking, automatic counting, DDK, oTMT, spontaneous speech) repeated-measures ANOVA revealed a significant main effect of gait complexity on step length (F = 41.63, p < 0.001, η²p = 0.094; indicating a medium effect size; online suppl. Table 1; for all online suppl. material, see https://doi.org/10.1159/000550799). Across conditions, step length was shorter during complex gait compared to simple gait (Fig. 2a). No significant main effect of cognitive/speaking task was observed (F = 0.90, p = 0.465, η²p = 0.009), and there was no interaction between gait complexity and cognitive/speaking task (F = 0.13, p = 0.970, η²p = 0.001).

Fig. 2.

Step length and dual-task effects (DTEs) in PwPD across gait complexity and cognitive/speaking tasks. Spaghetti plots show individual participants’ performance during simple gait (blue) and complex gait (orange). Step length (a) is shown across five cognitive/speaking tasks (2 × 5 ANOVA): single-task walking (ST Walking), automatic counting (Auto), diadochokinetic (DDK) task, Oral Trail Making Test (oTMT), and spontaneous speech (Spont). Motor dual-task effect (mDTE; b), cognitive dual-task effect (cogDTE; c), combined dual-task effect (cDTE; d), and attention allocation index (mAAI; e) are shown across four cognitive/speaking tasks used for DTE calculations (2 × 4 ANOVAs: Auto, DDK, oTMT, Spont). Significance values for main effects and interactions are indicated within each panel.

mDTE: the 2 (complexity) × 4 (cognitive/speaking task) ANOVA for mDTE demonstrated a significant main effect of cognitive/speaking task (F = 3.47, p = 0.016, η²p = 0.032; indicating a small-to-medium effect size; online suppl. Table 2). Post hoc pairwise comparisons indicated that performance during automatic counting differed significantly from spontaneous speech (t = 3.02, p_Bonferroni = 0.016, d = 0.47), with more DT interference during spontaneous speaking (online suppl. Table 3). Neither the main effect of gait complexity (F = 3.01, p = 0.084, η²p = 0.009) nor the interaction (F = 0.37, p = 0.771, η²p = 0.004) reached significance (Fig. 2b).

cogDTE: for cogDTE, there were no significant main effects of gait complexity (F = 0.90, p = 0.345, η²p = 0.003) or cognitive/speaking task (F = 2.34, p = 0.073, η²p = 0.022) and no interaction (F = 0.36, p = 0.781, η²p = 0.003; online suppl. Table 4; Fig. 2c).

cDTE: the ANOVA for cDTE revealed a significant main effect of cognitive/speaking task (F = 3.65, p = 0.013, η²p = 0.033; online suppl. Table 5). Post hoc pairwise comparisons indicated that performance during spontaneous speech differed significantly from oTMT (t = 2.96, p_Bonferroni = 0.020, d = 0.46), with more DT interference during spontaneous speaking (online suppl. Table 6). Neither gait complexity (F = 3.47, p = 0.063, η²p = 0.011) nor the interaction (F = 0.20, p = 0.897, η²p = 0.002) reached significance (Fig. 2d).

mAAI: the 2 × 4 ANOVA for mAAI showed a significant main effect of cognitive/speaking task (F = 3.36, p = 0.019, η²p = 0.031; indicating a small-to-medium effect size; online suppl. Table 7). Post hoc comparisons indicated more negative mAAI values during oTMT compared to automatic counting (t = 2.96, p_Bonferroni = 0.020, d = 0.46; online suppl. Table 8), indicating more attention allocation toward the cognitive/speaking task rather than walking. No significant main effect of gait complexity (F = 0.05, p = 0.824, η²p < 0.001) or interaction (F = 0.48, p = 0.695, η²p = 0.005) was observed (Fig. 2e).

Healthy Older Adult Group

Step length raw values: a 2 (gait complexity: simple, complex) × 5 (cognitive/speaking task: ST walking, automatic counting, DDK, oTMT, spontaneous speech) repeated-measures ANOVA revealed no significant main effects of gait complexity (F = 1.71, p = 0.195, η²p = 0.017) or cognitive/speaking task (F = 0.06, p = 0.993, η²p = 0.002) and no interaction (F = 0.04, p = 0.996, η²p = 0.002; online suppl. Table 9). Step length was similar across gait complexity and cognitive/speaking tasks (Fig. 3a).

Fig. 3.

Step length and dual-task effects (DTEs) in healthy older adults across gait complexity and cognitive/speaking tasks. Spaghetti plots show individual participants’ performance during simple gait (blue) and complex gait (orange). Step length (a) is shown across five cognitive/speaking tasks (2 × 5 ANOVA): single-task walking (ST Walking), automatic counting (Auto), diadochokinetic (DDK) task, Oral Trail Making Test (oTMT), and spontaneous speech (Spont). Motor dual-task effect (mDTE; b), cognitive dual-task effect (cogDTE; c), combined dual-task effect (cDTE; d), and attention allocation index (mAAI; e) are shown across four cognitive/speaking tasks used for DTE calculations (2 × 4 ANOVAs: Auto, DDK, oTMT, Spont). Significance values for main effects and interactions are indicated within each panel.

mDTE: the 2 (complexity) × 4 (cognitive/speaking task) ANOVA showed no significant main effects of gait complexity (F = 0.98, p = 0.326, η²p = 0.012) or cognitive/speaking task (F = 0.58, p = 0.632, η²p = 0.021) and no interaction (F = 0.63, p = 0.598, η²p = 0.023; online suppl. Table 10). mDTE values remained stable across gait complexities and cognitive/speaking tasks (Fig. 3b).

cogDTE: the ANOVA revealed a significant main effect of cognitive/speaking task (F = 4.99, p = 0.003, η²p = 0.158; indicating a large effect size; online suppl. Table 11). Post hoc analyses showed that cogDTE was significantly worse during automatic counting compared with DDK (t = 3.202, p_Bonferroni = 0.012, d = −0.97) and oTMT (t = 3.190, p_Bonferroni = 0.012, d = −0.96; online suppl. Table 12). No other pairwise comparison survived correction. Neither the main effect of gait complexity (F = 0.36, p = 0.552, η²p = 0.004) nor the interaction (F = 0.16, p = 0.923, η²p = 0.006) was significant (Fig. 3c).

cDTE: for cDTE, there was a significant main effect of cognitive/speaking task (F = 5.12, p = 0.003, η²p = 0.161; indicating a large effect size; online suppl. Table 13). Post hoc comparisons indicated that both DDK (t = −3.385, p_Bonferroni = 0.007, d = 1.02) and oTMT (t = −3.198, p_Bonferroni = 0.012, d = 0.96) showed significantly greater DT interference (more negative cDTE) compared with automatic counting, while no other pairwise comparisons were significant (online suppl. Table 14). Neither gait complexity (F = 0.0001, p = 0.993, η²p < 0.001) nor the interaction (F = 0.07, p = 0.974, η²p = 0.003) was significant (Fig. 3d).

mAAI: the ANOVA for mAAI revealed a significant main effect of cognitive/speaking task (F = 4.76, p = 0.004, η²p = 0.152; indicating a large effect size; online suppl. Table 15). Post hoc comparisons showed that oTMT (t = 3.258, p_Bonferroni = 0.010, d = 0.98) and DDK (t = 3.121, p_Bonferroni = 0.014, d = 0.95) were both significantly more negative than automatic counting, indicating greater attention allocation toward the cognitive/speaking task, while no other contrasts survived correction (online suppl. Table 16). No main effect of gait complexity (F = 0.83, p = 0.366, η²p = 0.010) or interaction (F = 0.30, p = 0.822, η²p = 0.011) was observed (Fig. 3e).

Step Length Coefficient of Variation Analyses

PwPD Group

Step length coefficient of variation raw values: a two-way repeated-measures ANOVA revealed a significant main effect of gait complexity (F = 280.42, p < 0.001, η²p = 0.412; indicating a large effect size; online suppl. Table 17), with greater step length coefficient of variation observed during complex compared to simple gait (Fig. 4a). There was no main effect of cognitive/speaking task (F = 0.84, p = 0.502, η²p = 0.008) and no interaction (F = 0.30, p = 0.880, η²p = 0.003).

Fig. 4.

Step length coefficient of variation (CV) and dual-task effects (DTEs) in PwPD across gait complexity and cognitive/speaking tasks. Spaghetti plots show individual participants’ performance during simple gait (blue) and complex gait (orange). Step length CV (a) is shown across five cognitive/speaking tasks (2 × 5 ANOVA): single-task walking (ST Walking), automatic counting (Auto), diadochokinetic (DDK) task, Oral Trail Making Test (oTMT), and spontaneous speech (Spont). Motor dual-task effect (mDTE; b), cognitive dual-task effect (cogDTE; c), combined dual-task effect (cDTE; d), and attention allocation index (mAAI; e) are shown across four cognitive/speaking tasks used for DTE calculations (2 × 4 ANOVAs: Auto, DDK, oTMT, Spont). Significance values for main effects and interactions are indicated within each panel.

mDTE: the 2 (complexity) × 4 (cognitive/speaking task) ANOVA revealed no significant main effects or interactions (Fig. 4b). Neither gait complexity (F = 3.23, p = 0.073, η²p = 0.010) nor cognitive/speaking task (F = 2.13, p = 0.097, η²p = 0.020) were significant, and the interaction was also nonsignificant (F = 0.61, p = 0.611, η²p = 0.006; online suppl. Table 18).

cogDTE: the 2 (complexity) × 4 (cognitive/speaking task) ANOVA revealed no significant main effects or interactions (Fig. 4c). Gait complexity was not significant (F = 0.90, p = 0.345, η²p = 0.003), cognitive/speaking task approached but did not reach significance (F = 2.34, p = 0.073, η²p = 0.022), and the interaction was nonsignificant (F = 0.36, p = 0.781, η²p = 0.003).

cDTE: the 2 (complexity) × 4 (cognitive/speaking task) ANOVA revealed no significant main effects or interactions (online suppl. Table 20). Neither gait complexity (F = 3.12, p = 0.078, η²p = 0.010), cognitive/speaking task (F = 1.03, p = 0.381, η²p = 0.010), nor the interaction (F = 0.03, p = 0.994, η²p < 0.001) reached significance (Fig. 4d).

mAAI: the 2 (complexity) × 4 (cognitive/speaking task) ANOVA revealed a significant main effect of cognitive/speaking task (F = 3.11, p = 0.027, η²p = 0.028; indicating a small-to-medium effect size; Fig. 4e; online suppl. Table 21). Post hoc pairwise comparisons did not identify any individual contrasts that reached statistical significance (online suppl. Table 22). Neither gait complexity (F = 1.07, p = 0.301, η²p = 0.003) nor the complexity × cognitive/speaking task interaction (F = 0.53, p = 0.664, η²p = 0.005) was significant.

Healthy Older Adult Group

Step length coefficient of variation raw values: the two-way repeated-measures ANOVA (online suppl. Table 23; Fig. 5a) revealed a significant main effect of gait complexity on step length coefficient of variation (F = 73.90, p < 0.001, η²p = 0.425; indicating a large effect size), with greater variability during complex walking. No significant main effect of cognitive/speaking task (F = 0.716, p = 0.58, η²p = 0.028) or interaction (F = 0.192, p = 0.942, η²p = 0.008) was found.

Fig. 5.

Step length coefficient of variation (CV) and dual-task effects (DTEs) in healthy older adults across gait complexity and cognitive/speaking tasks. Spaghetti plots show individual participants’ performance during simple gait (blue) and complex gait (orange). Step length coefficient of variation (a) is shown across five cognitive/speaking tasks (2 × 5 ANOVA): single-task walking (ST Walking), automatic counting (Auto), diadochokinetic (DDK) task, Oral Trail Making Test (oTMT), and spontaneous speech (Spont). Motor dual-task effect (mDTE; b), cognitive dual-task effect (cogDTE; c), combined dual-task effect (cDTE; d), and attention allocation index (mAAI; e) are shown across four cognitive/speaking tasks used for DTE calculations (2 × 4 ANOVAs: Auto, DDK, oTMT, Spont). Significance values for main effects and interactions are indicated within each panel.

mDTE: no significant main effects or interactions were observed for mDTE (online suppl. Table 24; Fig. 5b). Specifically, neither gait complexity (F = 1.34, p = 0.251, η²p = 0.016) nor cognitive/speaking task condition (F = 0.98, p = 0.404, η²p = 0.036) was significant, and the complexity × cognitive/speaking task interaction was also nonsignificant (F = 0.40, p = 0.752, η²p = 0.015).

cogDTE: the ANOVA revealed a significant main effect of cognitive/speaking task (F = 4.99, p = 0.003, η²p = 0.158; indicating a large effect size; online suppl. Table 25; Fig. 5c) but no significant effects of gait complexity (F = 0.36, p = 0.552, η²p = 0.004) or the complexity × cognitive/speaking task interaction (F = 0.16, p = 0.922, η²p = 0.006). Post hoc comparisons indicated worse cogDTE in the automatic counting condition compared with both DDK (t = −3.20, p_Bonferroni = 0.012, d = −0.97) and oTMT (t = −3.19, p_Bonferroni = 0.012, d = −0.96), while no other pairwise contrasts reached significance (online suppl. Table 26).

cDTE: no significant main effects or interactions were found for cDTE (online suppl. Table 27; Fig. 5d). Specifically, neither gait complexity (F = 1.69, p = 0.198, η²p = 0.021) nor cognitive/speaking task condition (F = 0.06, p = 0.980, η²p = 0.002) reached significance, and the complexity × cognitive/speaking task interaction was also nonsignificant (F = 0.29, p = 0.833, η²p = 0.011).

mAAI: the ANOVA revealed a significant main effect of cognitive/speaking task on mAAI (F = 3.25, p = 0.026, η²p = 0.109; indicating a medium-to-large effect size; online suppl. Table 28, Fig. 5e). Post hoc comparisons indicated that automatic counting resulted in significantly more positive mAAI values compared with oTMT (t = 2.80, p_Bonferroni = 0.038, d = 0.85), suggesting greater attentional allocation toward the motor task/walking during automatic counting. No other pairwise contrasts were significant (online suppl. Table 29).

PD and Healthy Older Adult Group Differences

Separate Welch’s ANOVAs were conducted to compare DTEs between Parkinson’s disease and healthy older adults for mDTE, cogDTE, cDTE, and mAAI, with group as the between-subjects factor. No significant differences between Parkinson’s disease and healthy older adults were observed for any of the DTE measures. As a result, pairwise comparisons were not performed. Specific numerical results and detailed statistics for each model are provided in online supplementary Tables 30 (for DTE metrics calculated with step length as the motor performance metric) 31 (for DTE metrics calculated with step length coefficient of variation as the motor performance metric).

Discussion

The purpose of this study was to explore differences in DT performance across various levels of cognitive and motor demands in PwPD and healthy older adults. Contrary to our expectations, there were no significant interaction effects between walking complexity and cognitive/speaking task. However, significant main effects were observed for both gait complexity (straight-line vs. complex walking) and cognitive/speaking task (oTMT, automatic counting, DDK, spontaneous speech). Overall, walking complexity influenced raw step length and step length coefficient of variation, particularly in PwPD, but did not change relative DTEs. Cognitive/speaking tasks had a greater influence on DTEs, with distinct patterns between groups. Contrary to our hypotheses, we observed no differences in DT performance between PwPD and healthy older adults.

This study addresses important gaps in the DT literature by simultaneously examining multiple cognitive/speaking tasks with distinct underlying demands across different types of gait complexity within a unified experimental paradigm. By comparing both simple and complex walking conditions and quantifying effects using raw gait metrics as well as DTEs, this work provides a more comprehensive characterization of cognitive-motor interference than studies that examine only a single cognitive task, a single walking condition, or gait outcomes alone [3, 6, 18]. This integrated approach allows clearer differentiation between changes in absolute gait performance and relative DT costs, thereby improving interpretation of how attentional demands and task characteristics shape DT performance in PwPD and healthy older adults.

Within-Group Effects of Gait Complexity

As hypothesized, gait complexity strongly influenced DT motor performance, particularly in PwPD. Both raw step length and step length coefficient of variation were significantly affected, with shorter and more variable steps observed during complex figure-eight walking with an obstacle compared to straight-line walking, consistent with prior work showing worse DT motor performance in PwPD during obstacle crossing [38]. This suggests that increased gait demands lead to shorter step length in PwPD, a characteristic gait deficit in Parkinson’s disease that progressively worsens with disease severity [39]. In healthy older adults, gait complexity increased step length variability while step length itself was preserved. This pattern suggests that older adults are able to maintain overall step size despite environmental demands but adapt by permitting greater variability. Importantly, this increased variability should not be interpreted as impaired motor control; rather, it may reflect a functional adaptation to meet the demands of navigating a more complex walking environment. Taken together, these findings indicate that complex gait paradigms challenge motor performance in both groups, with PwPD showing more pronounced effects, particularly in step length. These findings are consistent with prior work showing that PwPD exhibit reduced ability to modify gait in response to instructions during complex walking tasks, whereas healthy older adults retain more adaptive strategies such as adjusting variability or speed depending on task demands [40]. Interestingly, despite these clear changes in motor performance, gait complexity did not significantly influence DTEs in either group. This suggests that the relative cost of performing a secondary task while walking remains consistent across simple and complex walking conditions. In other words, participants appeared to adapt to DT demands similarly across walking conditions, despite the increased complexity in the walking condition.

Within-Group Effects of Cognitive/Speaking Tasks

The influence of cognitive/speaking tasks was more variable. In PwPD, for DTEs calculated with step length, cognitive/speaking task condition significantly affected mDTE, cDTE, and mAAI. Specifically, there was decreased DT capacity while walking with spontaneous speech compared to walking with oTMT. This may suggest that generative, spontaneous speech elicits greater demands relative to oTMT in PwPD, even though oTMT is thought to require higher attentional demand. Interestingly, post hoc analyses showed greater cognitive task prioritization (more negative mAAI) during walking with oTMT compared to walking with automatic counting. This shift in attention may help preserve overall DT performance. In contrast, during walking with spontaneous speech, overall DT performance decreases, suggesting the potential role of task structure and cognitive demand in promoting effective cognitive-motor coordination. These findings are consistent with evidence linking DT ability to cognitive function in PwPD [3, 15, 41, 42], and they further suggest that differences in performance across types and difficulties of cognitive/speaking tasks may highlight specific cognitive deficits that influence DT capacity in this population. Interestingly, cognitive/speaking tasks did not significantly affect raw step length or step length coefficient of variation, consistent with prior work showing that while DT generally impacts spatiotemporal gait parameters, the type of cognitive/speaking task does not differentially influence these gait measures [43]. However, another study reported that spatiotemporal gait parameters respond differently under DT, with serial subtraction by 7 producing the greatest gait interference, highlighting that the specific cognitive task can sometimes differentially affect gait [44]. Future studies should continue to explore how different cognitive/speaking tasks influence both motor and cognitive aspects of DT performance in PwPD, to better identify task-specific deficits and optimize rehabilitation strategies.

In healthy older adults, cognitive/speaking tasks had different effects, with significant effects of cognitive/speaking task on cogDTE, cDTE, and mAAI for DTEs calculated with step length, as well as cogDTE and mAAI calculated with step length coefficient of variation. Unlike PwPD, who did show effects in motor DTEs but not cognitive DTEs, healthy older adults showed significant effects in cognitive DTEs but not motor DTEs. Specifically, the rate of automatic counting declined during DT walking compared to walking with DDK or oTMT in healthy older adults. This is somewhat counterintuitive as the automatic counting task would seemingly require less attentional demand. However, in the DT context, performance in automatic counting declined in healthy older adults, suggesting that automatic or overlearned tasks may be more vulnerable to interference when paired with motor demands in this population. Additionally, healthy older adults demonstrated greater overall DT costs (cDTE) in DT with an automatic counting task compared to the other cognitive/speaking tasks, reinforcing this pattern.

Post hoc analyses of mAAI in healthy older adults further showed that walking with oTMT and DDK tasks resulted in more cognitive prioritization compared to walking with automatic counting, which consistently showed less cognitive prioritization. Importantly, this same pattern was seen in PwPD, where walking with automatic counting also showed less cognitive prioritization compared to walking with oTMT. Taken together, these findings suggest that walking with automatic counting may not elicit strong attentional allocation toward the counting task in either group, possibly because it is overlearned and requires little deliberate cognitive effort. In contrast, more complex and structured tasks, such as oTMT in PwPD and oTMT and DDK in healthy older adults, appear to demand more deliberate cognitive prioritization, potentially due to the nature of the task. Thus, automatic counting may act as a low-engagement cognitive condition, producing less prioritization across groups while still disrupting DT performance, especially in healthy older adults, when paired with walking.

Between-Group Comparisons of DTEs

Despite clear within-group effects of gait complexity and cognitive/speaking tasks, no significant differences in DTEs were observed between PwPD and healthy older adults across gait conditions or cognitive/speaking tasks. This contrasts with previous literature, which typically reports greater DT impairments in PwPD compared to healthy older adults [3, 45–47]. As shown in Table 1, our PwPD sample had a short disease duration, mild disease severity (MDS-UPDRS III and Hoehn and Yahr), and high functional capacity as evidenced by Mini-BESTest scores. These characteristics likely explain the comparable DTEs between groups as early-stage PwPD may retain sufficient motor automaticity and cognitive-motor integration to maintain relative DT performance similar to healthy older adults.

Limitations and Future Directions

Several limitations should be considered when interpreting the findings of this study. The sample size was modest, which may have limited statistical power to detect subtle between-group differences in DTEs, particularly given the variability inherent in cognitive-motor performance. In addition, participants with Parkinson’s disease were primarily in the early stages of the disease and demonstrated relatively mild motor impairment and high functional capacity. While this enhances internal consistency, it may limit generalizability to individuals with more advanced disease or greater mobility impairment. Task administration procedures also warrant consideration. Cognitive/speaking tasks were presented in a fixed order, and walking complexity condition order was not counterbalanced across participants; therefore, potential order, learning, or adaptation effects cannot be fully excluded. Testing was completed within a single session, which further precluded assessment of fatigue-related effects or longer-term adaptation to DT demands. Another methodological consideration relates to the automatic counting task. Because this task is highly overlearned and was performed without an externally constrained pace, participants may have produced unusually high response rates during ST conditions, potentially introducing ceiling effects. As a result, declines in performance during DT walking may reflect a return toward a more typical speaking pace rather than a true reduction in cognitive capacity. Although externally imposed pacing could reduce this effect, such constraints may also alter natural speech production and attentional demands. This highlights an important consideration for interpreting DTEs derived from unconstrained, overlearned tasks. Finally, the cognitive tasks included in this study were limited to speech-based measures, and findings may not generalize to other types of cognitive DTs (e.g., visuospatial or working memory tasks). Although the complex gait task incorporated obstacle negotiation and directional changes to better approximate environmental demands, it may not fully capture the complexity of real-world walking. Future studies should include larger and more diverse samples, incorporate a broader range of cognitive task modalities, and consider pacing constraints or normalization strategies to improve interpretability of DT costs. Longitudinal designs may also help clarify how DT performance evolves across disease progression and aging.

Conclusion

In this study, gait complexity affected step length and step length variability in PwPD, resulting in shorter and more variable steps during complex walking compared to straight-line walking. DTEs in PwPD were primarily influenced by cognitive/speaking tasks, with reduced capacity observed during walking with spontaneous speech compared to walking with oTMT. In healthy older adults, DT performance was also modulated by cognitive/speaking tasks, with greater cognitive costs observed during automatic counting compared to other cognitive/speaking task. These findings underscore the value of assessing both raw gait metrics and DT performance to capture the nuanced effects of motor and cognitive demands. Incorporating a range of motor and cognitive challenges may improve sensitivity in evaluating gait and DT performance, providing a more comprehensive understanding of automaticity and mobility in both early-stage PwPD and healthy older adults.

Acknowledgments

The authors sincerely thank all participants for their time, effort, and commitment to this study. Their valuable contributions made this research possible.

Statement of Ethics

This study protocol was reviewed and approved by the Saint Louis University Institutional Review Board (IRB #32321). Written informed consent was obtained from all participants prior to their participation in the study.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This work was supported by the Health Research Council at Saint Louis University, the Irma Ruebling Endowed Research Fund, the Parkinson’s Research and Support Fund, and the St. Louis Community Foundation. The funders had no role in the design, data collection, data analysis, or reporting of this study.

Author Contributions

S.J.C. contributed to conceptualization, data curation, formal analysis, project administration, visualization, investigation, writing – original draft, and writing – review and editing. E.B. contributed to investigation, writing – original draft, and writing – review and editing. J.H. contributed to investigation, data curation, and writing – review and editing. Y.-F.C. contributed to methodology, funding acquisition, conceptualization, project administration, supervision, and writing – review and editing. J.K.L. contributed to conceptualization, methodology, investigation, funding acquisition, project administration, supervision, and writing – review and editing.

Funding Statement

This work was supported by the Health Research Council at Saint Louis University, the Irma Ruebling Endowed Research Fund, the Parkinson’s Research and Support Fund, and the St. Louis Community Foundation. The funders had no role in the design, data collection, data analysis, or reporting of this study.

Data Availability Statement

The datasets generated and analyzed during the current study are not publicly available in order to protect participant confidentiality but may be made available to individuals upon reasonable request. Further inquiries can be directed to the corresponding author.