Effects of situational and social factors on locomotor behavior in chronic non-specific low back pain patients walking through apertures

Abstract

Chronic non-specific low back pain (cNSLBP) is a leading cause of disability. However, its impact on daily activities (e.g., walking in public spaces) remains poorly understood. This study aimed to examine how cNSLBP individuals make navigation decisions when exposed to varying situational and social conditions. Eighteen cNSLBP (45.7 ± 9.2 years, 10 women) and eighteen asymptomatic adults (AA, 43.7 ± 9.5 years, 7 women) walked a 10-meter path to reach a goal and chose to pass between two apertures placed midway according to four conditions: (1) baseline: two large apertures, (2) situational: narrow and large apertures, (3) situational and social: human interferer facing the narrow or for (4) large aperture. Measures included switch point, walking speed, clearance distance, and pain perception. Results revealed that cNSLBP participants adopted more conservative decision-making strategies. Under situational and social conditions, they showed a marked preference for the larger aperture (switch points per conditions (1) 0.1; (2) 1.83; (3) 3.35, and (4) 3.34), with weaker modulation of their choices based on goal position (p < 0.01) than AA participants (switch points per conditions (1) 0.2; (2) 0.70; (3) 1.77, and (4) 2.15). While all individuals adapted their strategy in the presence of a human interferer, cNSLBP participants gave less consideration to social norms, with no difference between switch points for conditions 3 and 4. In terms of kinematics, cNSLBP participants walked significantly slower than AA, regardless of condition or goal position (p < 0.001). For both groups, clearance distance significantly increased for all trials involving a human interferer (p < 0.05). No significant differences were observed between groups or conditions in shoulder rotation or in trunk torsion (p > 0.05). This paradigm offers a promising approach to addressing cNSLBP-related functional impairments, though further research is needed to clarify the underlying mechanisms.

Introduction

Low back pain (LBP) is one of the most prevalent types of pain and the leading cause of disability worldwide^1,2. It is defined as pain or discomfort located below the costal margin and above the lower gluteal folds, with or without leg pain¹. LBP is categorized as specific, neuropathic, or non-specific, with the latter accounting for the majority of cases³. While 70% of cases resolve within four weeks, LBP persisting beyond 12 weeks is classified as chronic⁴, and 85% of chronic cases are considered non-specific (chronic non-specific low-back pain, cNSLBP)¹. Clinical assessment of cNSLBP often relies on self-report scales and questionnaires evaluating pain intensity, emotional and cognitive aspects, and impact on daily activities^5,6. However, these tools do not fully capture how individuals interact with their environment, an essential dimension for understanding daily life functional difficulties. This limitation is increasingly highlighted by recent enactive and affordance-based models of pain, which propose that pain-related behavior is not solely driven by nociceptive input but is actively shaped by individuals’ perception of, and interaction with, their environment⁷.

The ecological approach to perception and action, particularly the concept of affordance⁸, emphasizes that perception is shaped by the actor-environment relationship, integrating both anthropometric and functional capacities^9,10,11. One widely used paradigm to study perceptual-motor behavior is walking through apertures, which reveals how individuals adapt their movements based on environmental constraints and body dimensions¹². The critical threshold, i.e., the aperture width at which individuals begin rotating their shoulders to pass, is typically observed between 1.3 and 1.4 times the shoulder width (Shoulder/Aperture ratio: S/A) in asymptomatic adults^10,11,13,14. However, critical threshold can increase in populations with specific characteristics. Older adults, for example, due to greater trunk sway and reduced walking speed, exhibit higher thresholds (1.6–1.7 S/A), correlated with their medio-lateral instability^9,15. Similar results were found in individuals with developmental coordination disorder¹⁶, and in children, whose increased variability in movement led to more conservative strategies¹⁷.

In this context, a recent study investigated the same task in cNSLBP individuals, and showed that symptomatic cNSLBP individuals rotated their shoulders at narrower thresholds (1.18 S/A) compared to asymptomatic individuals (1.33 S/A) ¹⁸. This suggests that cNSLBP patients adopt a risk-tolerant motor strategy, minimizing shoulder rotations to reduce stress on the painful lumbar region. Indeed, during walking, cNSLBP individuals tend to show increased muscle activity, reduced segmental mobility, and stiffer coordination patterns, reflecting greater rigidity^19,20. According to Hodges and Tucker²¹, such adaptations may serve as a protective mechanism, where pain or fear of pain triggers a reorganization of motor control aimed at lowering mechanical stress on the affected area. Another key finding in the aperture crossing task is that emotional, cognitive, or behavioral factors do not appear to influence the critical threshold in cNSLBP individuals¹⁸. This aligns with existing literature showing that fear of movement is linked to more cautious behavior in tasks like trunk flexion, which places greater demands on the painful lumbar region than simple walking tasks²². From this perspective, it seems important to explore more complex walking tasks, those involving interaction with the environment and requiring participants to use their trunk in a way similar to mobility tasks.

The aperture crossing paradigm has been extended to various situations by modifying the environment, for example, by narrowing the walking path or elevating it to increase postural threat²³, allowing participants to freely navigate around the aperture in an unconfined space⁹, or by incorporating social factors such as the presence of a person on the path^24,25. These adaptations show that, when faced with increased postural threat, individuals reduce their walking speed and exhibit greater body sway, while initiating shoulder rotation earlier, resulting in a higher critical threshold and a more cautious crossing strategy²³. In unconfined spaces, participants alter their behavior when the aperture is less than 1.4 S/A, preferring to circumvent the obstacle rather than pass through it, suggesting that bodily parameters remain key references even without strict spatial constraints, likely to preserve stability and momentum⁹. Furthermore, the introduction of human obstacles engages social mechanisms related to proxemics^25,26,27,28: the management of personal space, which is anisotropic and flexible, influences path selection, leading to larger safety margins and reduced speed when avoiding a person compared to an inanimate object. Although interest in perceptual-motor adaptations in cNSLBP individuals is growing, few studies have investigated their behavior in complex and dynamic environments. To address this gap, the present study examines the navigation strategies of individuals with cNSLBP during a task involving both situational and social constraints. This type of paradigm offers a realistic simulation of daily-life scenarios and may help to reveal specific adaptive difficulties in this population, while also extending research grounded in enactive models of pain⁷.

The study aims to examine whether cNSLBP influences decision-making when choosing a path between two apertures of different widths, depending on the presence of a human interferer (condition) and the goal’s position. We hypothesize that: (H1) in a baseline condition with two wide apertures, cNSLBP and asymptomatic adults (AA) will adopt similar strategies, choosing the aperture aligned with the goal (main effect of goal position); and (H2) when apertures differ in width and a human interferer is present, all participants will favor the wide aperture, even if the goal is behind the narrow one. This preference is expected to be stronger in cNSLBP participants, amplified by the presence of an interferer, and further enhanced when the interferer faces the narrow aperture (interaction effect between group, condition and goal position). Secondary objectives are twofold: first, to analyze gait kinematics underlying decision-making, and second, to evaluate the influence of pain perception on crossing strategies. We hypothesize that: (H3) participants will walk more slowly and show greater lateral clearance in the presence of an interferer, especially when facing the crossed aperture, with effects more pronounced in cNSLBP participants, who may also exhibit reduced shoulder rotation and less trunk torsion (interaction effect between group and condition); and (H4) trials in which the large aperture is chosen will correlate with dimensions of pain.

Results

Demographics

Age, sex and Mini Mental State Examination (MMSE) score were not significantly different between the cNSLBP and AA groups (p = 0.54, p = 0.50, and p = 0.20). The AA group had a higher quality of life score (W = 305, p < 0.0001, r = 0.76), a lower anxiety score (t(36)=-2.48, p < 0.01, d=-0.83) and a lower depression score (t(36)=-2.64, p < 0.01, d=-0.98) compared to the cNSLBP participants (Table 1).

Table 1 Participants’ characteristics.

Decision making

Decision-making is characterized by the switch point, defined in “Switch point”, which refers to the spatial threshold at which participants chose one aperture over the other, depending on the goal position.

Baseline condition

The baseline condition corresponds to the symmetrical condition (‘Sym’ condition, Fig. 1.A.), in which no situational or social factors are modified, i.e., with three poles creating two equal gaps (1.8 S/A). The goal labels on the left/right have been named ‘outside large left/right’ (2.7 S/A), ‘large left/right’ (1.8 S/A), ‘middle large left/right (0.9 S/A), and ‘center’ (0 S/A).

Fig. 1

Experimental set-up of walking task toward one of seven goals positioned 5 m beyond two apertures, under four conditions: (A) symmetrical (baseline): two large apertures (1.8 Shoulder / Aperture width – S/A), (B) asymmetrical (situational): one large and one narrow aperture (1.8 and 1 S/A), and situational and social: two unequal gaps created by poles and a human interferer positioned in the middle, facing either (C) the narrow and (D) large aperture. The goal positions were determined based on participants’ shoulder width and aperture width, placed at ratios of 2.7, 1.8 and 0.9 for the large aperture, and 0.5, 1 and 1.5 for the narrow aperture.

The logit-linked generalized linear mixed model (GLMM) revealed significant main effects of goal (OR = 0.018, 95%CI=[0.006, 0.052], p < 0.001), goal × group interaction (OR = 12.31, 95%CI=[4.09, 37.04], p < 0.001), but no main effect of group (OR = 0.93, p = 0.78). Post hoc pairwise comparison of estimated marginal trends (Bonferroni correction) showed that the effect of goal on aperture selection was significantly stronger in AA participants than in cNSLBP participants (β=-2.51 ± 0.56, z=-4.47, p < 0.0001). Although both groups switched apertures at similar widths (0.02 for AA, 0.01 for cNSLBP; Fig. 2.A.; supplementary materials 1), the AA group exhibited a steeper slope, indicating greater sensitivity to goal alignment.

Fig. 2

Switch point and slope estimates from the GLMM for each group and condition: Panel (A) symmetrical, (B) asymmetrical, (C) asymmetrical condition with human interferer facing the narrow gap, and (D) asymmetrical condition with human interferer facing the large gap. Below each panel, the switch point is indicated for each group, representing the goal position at which the probability of selecting either aperture reaches 50%. AA corresponded to Asymptomatic Adult and cNSLBP to chronic Non-Specific Low Back Pain.

Experimental conditions

Three experimental conditions were designed to manipulate situational and social factors (Fig. 1. B, C and D): the ‘Asym’ condition involved only situational asymmetry (a narrow gap of 1 S/A and a large gap of 1.8 S/A between three poles), the ‘HumNarrow’ condition combined situational and social factors, with a human interferer positioned in front of the narrow gap, and the ‘HumLarge’ condition replicated the same setup but with the interferer facing the large gap. Goal labels were defined ‘outside large’ (2.7 S/A), ‘large’ (1.8 S/A), ‘middle large (0.9 S/A) for the goal aligned with large aperture, and ‘middle small’ (0.5 S/A), ‘small’ (1 S/A), and outside small (1.5 S/A) for the goal aligned with narrow aperture. ‘Center’ (0 S/A) indicates the goal positioned at the center.

The logit-linked GLMM revealed a significant main effect of goal (OR = 0.002, 95%CI[0.1, 0.01], p < 0.001), but no main effect of group (p = 0.14). Significant effects were found for both the ‘HumNarrow’ (OR = 0.26, 95%CI[0.11, 0.60], p < 0.01) and ‘HumLarge’ (OR = 0.26, 95%CI[0.11, 0.62], p < 0.01) conditions. Several interactions were also significant: goal × group (OR = 97.91, 95%CI[28.12, 340.96], p < 0.001), goal × ‘HumNarrow’ (OR = 92.15, 95%CI[26.74, 317.53], p < 0.001), and goal × ‘HumLarge’ (OR = 91.42, 95%CI[26.50, 315.34], p < 0.001). Additionally, group × condition interactions were observed for ‘HumNarrow’ (OR = 5.88, 95%CI[2.26, 15.27], p < 0.001) and ‘HumLarge’ (OR = 6.96, 95%CI[2.65, 18.28], p < 0.001). Three-way interactions (goal × group × condition) were also significant for both ‘HumNarrow’ (OR = 0.02, 95%CI[0.01, 0.07], p < 0.001) and ‘HumLarge’ (OR = 0.02, 95%CI[0.00, 0.06], p < 0.001). The estimated slopes for each group and condition are presented in Fig. 2 (B, C and D) and supplementary materials 1.

• Post hoc pairwise comparisons of estimated marginal trends (Bonferroni correction) comparing groups within each condition showed that the relationship between goal and aperture selection was significantly steeper in the AA group than in the cNSLBP group for both the ‘Asym’ condition (β=-4.58 ± 0.64, z=-7.20, p < 0.0001) and the ‘HumNarrow’ condition (β=-0.60 ± 0.22, z=-2.70, p < 0.01). No significant difference between groups was found in the ‘HumLarge’ condition (p = 0.07).

• Post hoc pairwise comparisons of estimated marginal trends (Bonferroni correction) comparing conditions within groups showed that in the AA group, the slope was significantly reduced in both the ‘HumNarrow’ and ‘HumLarge’ conditions compared to the ‘Asym’ condition (β=-4.52 ± 0.63, z=-7.17, p < 0.0001 for both). No significant difference was observed between the ‘HumNarrow’ and ‘HumLarge’ conditions (p = 1.00). In the cNSLBP group, only the comparison between ‘Asym’ and ‘HumNarrow’ showed a significant slope reduction (β=-0.53 ± 0.21, z=-2.60, p < 0.05). No significant differences were found between ‘Asym’ and ‘HumLarge’ (p = 0.27), nor between ‘HumNarrow’ and ‘HumLarge’ (p = 0.50).

Switch points further illustrate these effects. In the ‘Asym’ condition (Fig. 2B), values were lower for AA (0.70) than cNSLBP (1.83). This gap increased in the ‘HumNarrow’ condition (Fig. 2C: 1.77 vs. 3.34) and remained in the ‘HumLarge’ condition (Fig. 2D: 2.15 vs. 3.35). Overall, cNSLBP participants consistently favored larger apertures, especially with a human interferer, whereas AA participants were more sensitive to goal position in the ‘Asym’ and ‘HumNarrow’ conditions.

Gait kinematics

Analyses were conducted separately for baseline and experimental conditions. For the experimental conditions, trials were analyzed based on whether participants passed through the large or narrow aperture, due to substantial imbalances in the number of observations across conditions and goals. Only goals for which a comparable number of participants crossed the respective aperture were included to ensure valid statistical comparisons, with exact numbers reported in Supplementary Materials 1–9. Specifically, when the large aperture was crossed, goals from ‘Outside Large’ to ‘Middle Small’ were analyzed, whereas for the narrow aperture, goals from ‘Middle Small’ to ‘Outside Small’ were included.

Walking speed

In baseline condition (Fig. 3A, supplementary materials 2), Mauchly’s test revealed a violation of sphericity for goal and the group × goal interaction [W = 0.26, p = 0.003], and so Greenhouse-Geisser (GG) correction was applied (ε = 0.697). A repeated measures ANOVA showed a significant main effect for group (F(34,1) = 19.60, p = 0.001, η²_p = 0.360) and goal (F(142.13,4.18) = 3.07, p = 0.001, η²_p = 0.002), but no significant group × goal interaction (p = 0.18). Post hoc pairwise t-tests (Bonferroni correction) revealed that cNSLBP participants walked significantly slower than AA participants across all goals (p < 0.0001).

Fig. 3

Kinematic data from the symmetrical condition. Panel (A) shows walking speed during aperture crossing. Panel (B) shows the clearance distance, expressed as a Clearance Distance/Aperture Ratio (C/A), during aperture crossing. The main effect of goal position is indicated by a horizontal bar. AA corresponds to Asymptomatic Adult and cNSLBP to chronic Non-Specific Low Back Pain. Bar plots represent the means, with error bars indicating the standard deviations.

In experimental conditions involving the large aperture (Fig. 4A, supplementary materials 4), Mauchly’s test revealed a violation of sphericity for condition and the group × condition interaction for the ‘outside large’ (W = 0.74, p < 0.01), ‘large’ (W = 0.62, p < 0.001), and ‘middle large’ (W = 0.70, p < 0.01) goals, therefore, and so GG correction was applied (ε = 0.792, ε = 0.726 and ε = 0.771). No violation was detected for the ‘center’ (W = 0.98, p = 0.77) and ‘middle small’ (W = 0.83, p = 0.16) goals. A mixed-design ANOVA showed a significant main effect of group for all goals (‘outside large’ (F(34,1) = 17.04, p < 0.001, η²ₚ=0.32), ‘large’ (F(34,1) = 15.48, p < 0.001, η²ₚ=0.30), ‘middle large’ (F(34,1) = 13.98, p < 0.001, η²ₚ=0.28), ‘center’ (F(34,1) = 15.14, p < 0.001, η²ₚ=0.30), and ‘middle small’ (F(20,1) = 10.44, p < 0.01, η²ₚ=0.34)). No significant main effect of condition (‘outside large’ (p = 0.45), ‘large’(p = 0.45) and ‘middle large’ (p = 0.39), ‘center ’ (p = 0.22) and ‘middle small’ (p = 0.11)) or group × condition interaction (‘outside large’ (p = 0.44), ‘large’(p = 0.72) and ‘middle large’ (p = 0.76), ‘center’ (p = 0.52) and ‘middle small’ (p = 0.29) was found. Post hoc pairwise t-tests (Bonferroni correction) revealed that cNSLBP participants walked significantly slower than AA participants across all goals (all p < 0.001).

Fig. 4

Kinematic data, in which participants selected a large aperture, from symmetrical condition (‘Sym’), asymmetrical condition (‘Asym’), asymmetrical condition with human interferer facing the narrow gap (‘HumNarrow’), and asymmetrical condition with human interferer facing the large gap (‘HumLarge’). Panel (A) shows walking speed during aperture crossing. Panel (B) shows the clearance distance, expressed as a Clearance Distance/Aperture Ratio, during aperture crossing. The interaction (group x condition) is illustrated as follow: a and b indicate significant differences within groups (AA and cNSLBP, respectively), c indicates significant difference between groups in the HumLarge condition, and d indicates main effect of condition. AA corresponds to Asymptomatic Adult and cNSLBP to chronic Non-Specific Low Back Pain. Bar plots represent the means, with error bars indicating the standard deviations.

In experimental conditions involving the narrow aperture (Fig. 5A, supplementary materials 6), no violation of the sphericity assumption was observed for the ‘middle small’ (W = 0.52, p = 0.10), ‘small’ (W = 0.71, p = 0.13), and ‘outside small’ (W = 0.77, p = 0.15) goals. A mixed-design ANOVA showed a significant main effect of group across all goals (‘middle small’ (F(1, 8) = 20.35, p < 0.01, η²ₚ=0.67), ‘small’ (F(1, 13) = 17.83, p < 0.001, η²ₚ=0.96), and ‘outside small’ (F(1, 16) = 21.81, p < 0.001, η²ₚ=0.96)). No significant effects of condition (‘middle small’: p = 0.88; ‘small’: p = 0.73; ‘outside small’: p = 0.39) or group × condition interaction (‘middle small’: p = 0.99; ‘small’: p = 0.39; ‘outside small’: p = 0.39) were observed. Post hoc pairwise t-tests (Bonferroni correction) revealed that cNSLBP participants walked significantly slower than AA participants across all goals (p < 0.0001).

Fig. 5

Kinematic data, in which participants selected a narrow aperture, from symmetrical condition (‘Sym’), asymmetrical condition (‘Asym’), asymmetrical condition with human interferer facing the narrow gap (‘HumNarrow’), and asymmetrical condition with human interferer facing the large gap (‘HumLarge’). Panel (A) shows walking speed during aperture crossing. Panel (B) shows the clearance distance, expressed as a Clearance Distance/Aperture Ratio, during aperture crossing. Panel (C) shows the shoulder rotation during aperture crossing. AA corresponds to Asymptomatic Adult and cNSLBP to chronic Non-Specific Low Back Pain. d indicates main effect of condition. AA corresponded to Asymptomatic Adult and cNSLBP to chronic Non-Specific Low Back Pain. Bar plots represent the means, with error bars indicating the standard deviations.

Clearance distance

In baseline condition (Fig. 3B, supplementary materials 3), Mauchly’s test revealed a violation of sphericity for goal and group × goal interaction (W = 0.04, p < 0.001), and so GG correction was applied (ε = 0.553). A repeated measures ANOVA showed a significant main effect of goal (F(112.8, 3.32) = 37.97, p < 0.0001, η²_p = 0.25), but neither the main effect of group (p = 0.15) nor the group × goal interaction (p = 0.34) reached significance. Post hoc pairwise t-tests (Bonferroni correction) showed that clearance distance (C/A) from the central pole increased as the goal was placed laterally: ‘outside large left’ (p < 0.05) > ‘large left’ (p < 0.05) > ‘middle large left’ (p < 0.05), and similarly, ‘outside large right’ (p < 0.05) > ‘large right’ (p < 0.05) > ‘middle large right’ (p < 0.05). No significant differences were observed between goal placed symmetrically on the left and right sides: ‘outside large left’/‘outside large right’ (p = 0.23), ‘large left’/‘large right’ (p = 0.71), and ‘middle large left’/‘middle large right’ (p = 0.95).

In experimental conditions involving the large aperture (Fig. 4B, supplementary materials 5), Mauchly’s test revealed a violation of sphericity for condition and the group × condition interaction for the ‘outside large’ (W = 0.94, p = 0.35), ‘large’ (W = 0.88, p = 0.12), ‘middle large’ (W = 0.86, p = 0.086), and ‘center’ (W = 0.89, p = 0.14) goals. However, for the ‘middle small’ goal, sphericity was violated (W = 0.50, p < 0.001), and so GG correction was applied (ε = 0.878).

• For the ‘outside large’ and ‘large’ goals, a mixed-design ANOVA showed a main effect of condition (F(2, 68) = 81.15, p < 0.001, η²ₚ=0.34; F(2, 68) = 74.06, p < 0.0001, η²ₚ=0.35) and a group × condition interaction (F(2, 68) = 3.48, p < 0.001, η²ₚ=0.02; F(2, 68) = 8.76, p < 0.001, η²ₚ=0.06). A follow-up ANOVA (Bonferroni correction) showed a significant group effect in the ‘HumLarge’ condition (F(1, 34) = 7.49, p < 0.05, η²=0.18; F(1, 34) = 15.8, p < 0.01, η²=0.32), with no difference in the ‘Asym’ (p = 0.49; p = 0.26) or ‘HumNarrow’ (p = 0.18, p = 0.08) conditions. Post hoc pairwise t-tests (Bonferroni correction) showed that cNSLBP participants exhibited greater clearance distances than AA participants in the ‘HumLarge’ condition (p < 0.01 in both). Within-group analyses revealed significant effects of condition in both AA (F(2,51) = 9.72, p < 0.001, η²=0.27; F(2, 51) = 11.1, p < 0.001, η²=0.30) and cNSLBP (F(2, 51) = 17.0, p < 0.001, η²=0.4; F(2, 51) = 18.3, p < 0.001, η²=0.42) groups. Post hoc pairwise t-tests (Bonferroni correction) showed that the clearance distances were greater in the ‘HumLarge’ and ‘HumNarrow’ conditions compared to the ‘Asym’ condition (p < 0.01 in both), with no significant difference between the two human-interferer conditions (AA: p = 0.86 and p = 0.43; cNSLBP: p = 1 and p = 0.96).

• For the ‘middle large’, ‘center’, and ‘middle small’ goals, a mixed-design ANOVA showed a main effects of condition (F(2, 68) = 117.98, p < 0.001, η²ₚ=0.43; F(2, 68) = 131.60, p < 0.001, η²ₚ=0.45; F(1.33, 26.61) = 101.70, p < 0.01, η²ₚ=0.90)) and group (F(1, 34) = 4.58, p < 0.001, η²ₚ=0.09; F(1, 34) = 8.03, p < 0.001, η²ₚ=0.16; F(1, 20) = 10.45, p < 0.01, η²ₚ=0.30)), without significant group × condition interactions (p = 0.12, p = 0.12, and p = 0.14). Post hoc pairwise t-tests (Bonferroni correction) showed that the clearance distances were greater in cNSLBP group than AA (p < 0.05, p < 0.01, and p < 0.001), and for both ‘HumLarge’ and ‘HumNarrow’ compared to ‘Asym’ (all p < 0.0001), with no significant difference between the two human-interferer conditions (p = 0.23, p = 0.21, and p = 0.50).

In experimental conditions involving the narrow aperture (Fig. 5B, supplementary materials 7), the assumption of sphericity was violated for the ‘small’ (W = 0.60, p = 0.05) and ‘outside small’ (W = 0.65, p = 0.04) goals, and so GG correction was applied (ε = 0.712 and ε = 0.74). For the ‘middle small’ goal (W = 0.84, p = 0.55), sphericity was not violated. A mixed-design ANOVA showed a significant main effect of condition for all goals (‘middle small’ (F(2, 16) = 12.13, p < 0.001, η²ₚ=0.31), ‘small’ (F(1.42, 18.52) = 20.20, p < 0.001, η²ₚ=0.27), and ‘outside small’ (F(2, 32) = 23.91, p < 0.001, η²ₚ=0.33)). No significant group effects or group × condition interactions were identified. Post hoc pairwise t-tests (Bonferroni correction) indicated that clearance distances were greater in the ‘HumNarrow’ and ‘HumLarge’ conditions than in ‘Asym’ for ‘middle small’ and ‘outside small’ goals (p < 0.01 and p < 0.0001). No differences were found between ‘HumNarrow’ and ‘HumLarge’ (p = 0.48 and p = 0.99). For the ‘small’ goal, the clearance distances were greater for ‘HumNarrow’ compared to ‘Asym’ condition (p < 0.01), with no significant differences between ‘Asym’ and ‘HumLarge’ (p = 0.07) or between ‘HumNarrow’ and ‘HumLarge’ (p = 0.11).

Shoulder rotation

In experimental conditions involving the narrow aperture (Fig. 5C, supplementary materials 8), the assumption of sphericity was met for all goals: ‘middle small’ (W = 0.99, p = 0.97), ‘small’ (W = 0.85, p = 0.38), and ‘outside small’ (W = 0.81, p = 0.21). A mixed-design ANOVA showed a significant main effect of condition for ‘middle small’ (F(2, 16) = 5.42, p < 0.05, η²ₚ=0.11), ‘small’ (F(2, 26) = 9.66, p < 0.001, η²ₚ=0.09), and ‘outside small’ (F(2, 32) = 8.42, p < 0.001, η²ₚ=0.09). No main effect of group (‘middle small’ (p = 0.45), ‘small’(p = 0.79) and ‘outside small’ (p = 0.56)) or group × condition interaction (‘middle small’ (p = 0.99), ‘small’(p = 0.48) and ‘outside small’ (p = 0.34)) were identified. Post hoc pairwise t-tests (Bonferroni correction) showed no significant differences in shoulder rotation for any goal between ‘Asym’ and ‘HumNarrow’ conditions (p = 0.14, p = 0.49 and p = 0.09), ‘Asym’ and ‘HumLarge’ (p = 1, p = 1, p = 0.16), and ‘HumNarrow’ and ‘HumLarge’ conditions (p = 0.79, p = 1, p = 1).

Trunk torsion

In experimental conditions involving the narrow aperture (supplementary materials 9), sphericity was violated for all goals: ‘middle small’ (W = 0.29, p = 0.01, ε = 0.585), ‘small’ (W = 0.55, p = 0.03, ε = 0.69), and ‘outside small’ (W = 0.53, p < 0.01, ε = 0.681). a mixed-design ANOVA showed no significant effects of condition (‘middle small’: p = 0.07; ‘small’: p0.11; ‘outside small’: p = 0.40), group (‘middle small’: p = 0.82; ‘small’: p = 0.52; ‘outside small’: p = 0.71) or group × condition interaction (p = 0.30, p = 0.46 and p = 0.67) were identified.

Pain perception

For each condition and goal, no significant correlation was found between variables related to pain perception and the percentage of cNSLBP participants passing the large aperture (p > 0.05). Scores from the RMDQ, FABQ, PCS, PIPS, TSK questionnaires and VAS are presented in Table 1. Percentages of large aperture crossings are reported in supplementary materials 10.

Discussion

The primary aim of the study was to determine whether decision-making during locomotion in individuals with cNSLBP is influenced by situational (aperture width) and social (presence of a human interferer) factors when choosing between two apertures.

The main outcome was the switch point, defined as the goal position at which participants selected an aperture. Consistent with H1, results in the ‘Sym’ condition revealed that both groups exhibited central switch points, suggesting comparable strategies of aligning their choice with the goal location. However, the AA group displayed a steeper slope, reflecting greater sensitivity to goal position. In line with H2, all participants favoured the large aperture, even when the goal was positioned beyond the narrow aperture. The participants’ preference for passing through larger apertures is consistent with previous evidence that apertures narrower than 1.4 S/A tend to be bypassed rather than passed through⁹. Fajen and Warren^29,30 proposed that such behavior reflects a trade-off between goal attraction and obstacle repulsion that would afford safe passage: obstacles forming apertures smaller than 1.4 S/A would create a repulsion of the aperture to be passed that would outweigh the attraction of the goal and force individuals to modify their trajectory⁹. However, a group difference emerged such that in the ‘Asym’ condition, AA had a lower switch point than cNSLBP and a steeper slope, showing stronger modulation by goal position. Under social constraints, switch points increased in both groups, particularly in cNSLBP, highlighting a stronger avoidance of the narrow aperture. Slopes decreased in both groups compared to ‘Asym,’ with a sharper reduction in cNSLBP, suggesting their path selection was less goal-driven and more cautious, avoidance-oriented choices in social contexts. Beyond situational constraints, the presence of a human interferer influenced behavior. Proxemics research^26,27,28 demonstrates participants rotate their shoulders despite a clear passage when the aperture is framed by humans rather than poles²⁴,or with increased avoidance when the human interferers face the path^25,31. Consistent with this finding, AA participants performed goal-sensitive strategies in the ‘Asym’ condition, but this modulation diminished with a human interferer. By contrast, cNSLBP participants only differed between ‘Asym’ and ‘HumNarrow’ conditions. Furthermore, AA adjusted their switch point more in ‘HumNarrow’ than ‘HumLarge,’ unlike cNSLBP, supporting the notion that AA integrate both social and situational cues, while cNSLBP rely less on social norms.

The secondary objectives of the study were to analyze participants’ gait kinematics (speed, clearance distance, shoulder rotation and trunk torsion) and to assess the influence of pain perception variables on the decision-making of aperture passed.

Contrary to H3, the presence of an interferer did not significantly affect walking speed, shoulder rotation, or trunk torsion, but it did influence collision avoidance. Concerning walking speed, the findings contrast previous findings, which typically reported slower walking speeds when navigating narrower apertures^14,16, allowing additional time for visuomotor planning³², directional changes³³, and collision avoidance¹⁴. Hackney et al. ²⁴ also observed slower speeds when participants passed between human interferers, with further reductions as aperture width decreased. However, the short 5-meter path in our study may have limited participants’ ability to reach or adjust a stable walking speed. Similar results have been reported in studies with shorter paths (e.g., 4 m), where neither aperture width nor interferer orientation significantly influenced walking speed²⁵. Concerning shoulder rotation, the absence of significant differences contrasts with earlier studies revealing greater rotation when apertures were formed by two humans facing the gap compared to when their backs were turned^24,25. This discrepancy likely reflects differences in experimental paradigms and aperture configurations, as our apertures were not defined by two individuals. However, clearance distance revealed important effects. Participants maintained greater lateral clearance when a human interferer was present compared to a configuration with poles, regardless of whether the aperture was large or narrow. These findings align with previous literature indicating that participants maintain larger clearance distances when passing near human interferers, across narrow apertures and various configurations²⁵. However, while clearance distance was adjusted according to situational and social factors, the orientation of the human interferer did not appear to influence movement strategy. Regarding group differences, significant effects were observed for walking speed and clearance distance. First, cNSLBP participants walked slower than AA participants across all goal positions and experimental conditions, consistent with previous findings²⁰. Second, differences in clearance distance emerged only for the most lateral goals on the wide side (‘outside large’ and ‘large’), with cNSLBP participants passing more laterally than AA participants. Finally, the absence of group differences in shoulder rotation can be explained by the fact that a 1 S/A aperture still requires shoulder rotation to pass¹⁸. All participants who traversed the narrow aperture rotated their shoulders, but only a small number of cNSLBP participants chose to do so. The lack of difference in shoulder rotation amplitude between AA and cNSLBP suggests that those who crossed were willing to reconfigure their bodies to fit through the narrow gap.

Regarding pain perception, contrary to H4, no significant influence of variables such as pain intensity, fear and avoidance beliefs, kinesiophobia, anxiety, depression, false beliefs, or quality of life was observed on the decision-making of passing a large aperture. One possible explanation is the use of generic questionnaires. According to some authors³², these tools may be too broad to capture task specifics and therefore they recommend using questionnaires that directly address the difficulties participants might encounter in a given task³². Further studies could therefore be carried out on the creation of specific questionnaires to measure fear of movement and beliefs as part of a study on the interaction between the individual and the environment²².

Finally, these results do not yet support immediate clinical application but align with explanatory models like the enactive model⁷, which views pain as an embodied, situated experience co-constructed through action. This model considers the patient as an active agent shaping their experience, with treatments focused on supporting meaning-making, interaction, and engagement. The navigation task, a daily activity, fits well within this framework. Future research should explore whether interventions targeting navigation strategies can improve the patient’s relationship to pain and if changes in navigation correlate with clinical improvement, aiding the development of assessment and monitoring tools centered on lived experience.

Limits

Our study has several limitations. First, regarding social factors, we used a single human interferer (a 30-year-old woman of average height) and couldn’t control for participant variables such as sex, age, and height. Due to the limited number of trials, we couldn’t further analyze their impact, although previous research conducted in laboratory settings^33,34,35,36 and virtual reality environments^{37,38,39,40,41,42} has demonstrated their relevance. Second, the experimental protocol was limited to a single scenario where participants knew the interferer would remain stationary, which may have influenced navigation behavior^26,27,43,44. Moreover, the study took place in a controlled indoor environment with one walker and one human interferer, limiting the applicability to more dynamic contexts involving multiple moving interferers or varying spatial constraints^41,43,44,45. Third, the generalizability of the results is constrained by the relatively low average pain intensity in our sample, as pain levels can fluctuate depending on the time of day or the participant’s condition, and it is possible that pain was reduced during the experimental session. Fourth, the lateralization of pain in cNSLBP participants was not considered during recruitment or analysis, which may have influenced decision-making. Fifth, the small sample size, although sufficient for between-group comparisons, may limit the ability to detect robust correlations between locomotor behaviors and pain-related clinical outcomes. Finally, while we followed recommended methods for assessing cNSLBP^5,6, our focus was on emotional aspects of pain, neglecting cognitive and social factors. Although few studies have examined the impact of cNSLBP on cognitive performance⁴⁶, existing findings may provide valuable insights. Due to session duration, additional clinical assessments were not included.

Conclusion

This study revealed differences in decision-making behaviors between cNSLBP and AA individuals exposed to environments changed by situational and social factors. While AA participants favored the shortest path, even though narrow apertures, cNSLBP participants preferred longer routes using larger gaps. Additionally, all individuals adjusted their strategy in the presence of a human interferer, however cNSLBP gave less consideration to social norms. Hence, under certain conditions, it may be suggested that chronic pain alters the relative weighting assigned to situational and social factors during navigation. Further research exploring different situations and adjusting parameters, is needed to better understand the role of situational and social factors in decision-making.

Methods

Participants

The study was approved by the Committee for the Protection of Individuals Île-de-France VI, France (2023-A00870-45) and procedures were performed in accordance with ethical guidelines (NCT06644053). Eighteen cNSLBP adults (45.7 ± 9.2 years, 10 females) and eighteen asymptomatic adults (AA, 43.7 ± 9.5 years, 7 females) volunteered to the experiment and provided written informed consent. The number of participants was determined based on previous studies using similar paradigms comparing two groups^9,16. AA participants were recruited through advertisements, and cNSLBP participants were recruited from three medical centers: SPORMED (a center for Sports Medicine and Effort Rehabilitation, Rennes, France), the Department of Physical Medicine and Rehabilitation at Rennes University Hospital (Rennes, France), and the Patis-Fraux center (Vern-sur-Seiche, France). Each participant’s inclusion was carried out by one of the two investigating physicians of the study (O.R. A.B. and P.C.J.). The cNSLBP adults had to suffer from low back pain for more than 12 weeks, at least 4 days a week and with a self-evaluated average pain ≥ 4/10 on Visual Analog Scale (VAS) on the day of inclusion, without initiating or modifying any analgesic treatment, except if it had been stable for a defined period before inclusion (e.g., ≥ 2 weeks). Exclusion criteria included pregnancy or breastfeeding, uncorrectable visual impairments, cognitive impairments, psychological or neurological disorders, or any clinical red flags suggestive of an underlying pathology requiring specific or urgent medical treatment. In addition, the AA group did not have ongoing chronic pain or a history of significant chronic pain (≥ 4/10 on VAS for at least 6 months) and pathologies that affect an individual’s ability to walk or maintain an optimal posture³.

Experimental design

The experiment was conducted along a 10 m path with two static door-like apertures positioned 5 m from the participants’ starting point, located in the middle of a gymnasium (20 × 30 m). A single goal was placed 5 m beyond the aperture in one of seven positions relative to the poles (see protocol section for specific goals). Each trial featured two apertures formed by three obstacles, two outer poles and a central obstacle, which was either a pole or a human interferer, set to widths of either 1.0 or 1.8 times each participant’s shoulder width (Fig. 1). The optoelectronic motion capture system (Qualisys, Gothenburg, Sweden), with 23 cameras operating at 200 Hz, was used to track the position of fourteen passive retroreflective markers over time, markers were placed on participants’ glenohumeral (GH) joints, the 7th cervical vertebrae, the anterior and posterior superior iliac spines (ASIS and PSIS), the external ankle malleoli and on a helmet (5 markers).

Protocol

First, participants completed self-reports on anxiety and depression (Hospital Anxiety Depression Scale, HADS⁴⁷ and on quality of life (EuroQol 5 Dimension, EQ-5D ⁴⁸). In the cNSLBP group, pain perception was assessed focusing on disability (RMDQ⁴⁹, fear avoidance belief (FABQ⁵⁰, pain catastrophizing (PCS⁵¹, kinesiophobia (TSK⁵², psychological inflexibility (PIPS⁵³ and pain intensity with a visual analogical scale (VAS: 0 ‘no pain’- 10 ‘worst imaginable pain’ ) described in Table 1.

Second, each participant’s shoulder width (measured as the distance between the most lateral points of the left and right GH joints) was recorded to determine their relative aperture size. We took into account the medio-lateral size of the pole and the participant to standardize the width of the aperture. Participants were asked to walk at a comfortable speed and reach one of the seven goals located behind the two apertures, choosing which aperture to pass through. Four experimental conditions were designed to manipulate situational and social factors (Fig. 1). The ‘Sym’ condition (baseline) involved no modifications to situational or social factors, with three poles creating two equal gaps of 1.8 S/A each, the ratio does not require movement adaptation for cNSLBP and AA participants¹⁸. The ‘Asym’ condition introduced only situational asymmetry, with a narrow gap of 1 S/A and a large gap of 1.8 S/A between the three poles. The ratio of 1 represents an aperture equal in width to the participant’s shoulders width, requiring movement adaptation, such as trunk rotation, to pass through without touching the aperture¹⁸. The ‘HumNarrow’ and ‘HumLarge’ conditions replicated this setup but included a human interferer positioned in the middle of the gaps, facing either the narrow or the large gap, respectively.

The goal positions were determined based on participants’ shoulder width and aperture width, placed at ratios of 2.7, 1.8 and 0.9 for the large aperture, and 0.5, 1 and 1.5 for the narrow aperture. In ‘Sym’ condition the goal labels on the left/right have been named ‘outside large left/right’ (2.7 S/A), ‘large left/right’ (1.8 S/A), and ‘middle large left/right (0.9 S/A). In the other conditions the goal labels have been named ‘outside large’ (2.7 S/A), ‘large’ (1.8 S/A), ‘middle large’ (0.9 S/A), ‘middle small’ (0.5 S/A), ‘small’ (1 S/A) and ‘outside small’ (1.5 S/A). In all conditions, the goal position at 0 S/A was labeled ‘center’.

Participants completed four conditions of 28 trials each. The order of the conditions was counterbalanced across participants, and the trials within each condition were randomized individually for each participant. Additionally, the trials from conditions 3 and 4, which involved the human interferer, were randomized and divided into two separate conditions to allow for a break every 28 trials.

At the start and the end of the four conditions, participants completed two 10 m trials to measure their comfortable walking speed in the absence of obstacles along their path. In addition, cNSLBP participants rated their pain on the VAS scale prior to the first condition and after each subsequent condition.

Data analysis

Data analysis was performed using MATLAB (Version R2021b, MathWorks, Natick, USA). To remove oscillations due to stepping activity, position data were filtered using a 10 Hz Butterworth low pass filter. We then computed the following kinematic variables:

Switch point

The position of each participant’s head center was tracked at each time point to analyze the switch point across experimental conditions (situational and social characteristics). The switch point was defined as the goal position at which participants changed their decision-making, specifically, the goal at which they chose to pass through the large aperture for the ‘Sym’ condition and the narrow aperture for the other three conditions. To determine the switch point for each condition, we conducted a logistic regression analysis predicting the probability that participants selected the narrow aperture (requiring movement adaptation) versus the large aperture (requiring no adaptation), based on the lateral position of the goal. The switch point was identified as the goal position corresponding to a 50% probability of choosing either aperture, representing the threshold where participants’ preference shifted from one aperture to the other¹².

Walking speed

Walking speed was computed using the first time-derivative of the participant’s position (the location of the center of each participant’s head at each time point). Walking speed corresponded to the walking velocity at the moment of passing through the aperture, that is, when participants crossed the Y-axis¹⁸ (value of 0).

Clearance distance

The clearance distance was expressed as a relative value by calculating the ratio between the clearance distance and the aperture width (C/A), where 0 corresponds to the position of the central obstacle between the two openings, and 1 to the position of one of the two lateral poles. This measure allowed us to determine the participant’s lateral position within the aperture at the moment of passage. The clearance distance was calculated based on the horizontal coordinate of the GH joint closest to the central obstacle.

Shoulder rotation

The shoulder rotation of each participant was calculated from the horizontal coordinates of the two GH joints. The angle was defined in the horizontal plane between the GH line and the frontal plane of the participants’ body if they were to be walking forward in an anatomically neutral position when participants passed the aperture. To compare the shoulder rotation of both groups, we used the absolute value of the angle (i.e., the magnitude of rotation).

Trunk torsion

Trunk torsion was defined as the angle between the scapular and pelvic belts, calculated from the horizontal coordinates of the two GH joints and the ASIS. The angle was defined in the horizontal plane between the GH line and the ASIS line when participants passed the aperture.

Statistics

Statistical analysis was performed using RStudio (version 4.1.3, RStudio, Inc). The level of significance was set to α = 0.05. Normality and homogeneity of variance were assessed using Shapiro-Wilk and Levene tests. For age and scores on the HADSA and HADSD questionnaires, the assumptions of normality and homogeneity of variances were met, therefore, Student’s t-tests were applied. In contrast, MMSE scores showed significant departures from normality in both groups (Shapiro-Wilk p < 0.01), although variances were equal (Levene’s test p = 0.623). For Eq. 5D scores, normality was violated in the AA group (p = 0.011), and a significant variance inequality was observed across groups (Levene’s test p = 0.002). Accordingly, non-parametric Wilcoxon rank-sum tests were used for both MMSE and Eq. 5D group comparisons. A Pearson’s Chi-squared test was conducted to examine sex distribution across groups. Given that all expected cell counts were greater than 5, the test assumptions were met.

To address our primary objective, determining whether decision-making during locomotion in individuals with cNSLBP is influenced by situational factors and social factors, we conducted the statistical analyses, separating the data from the ‘Sym’ condition from those of the experimental conditions (‘Asym’, ‘HumNarrow’, and ‘HumLarge’ conditions), due to differences in goal position.

• ‘Sym’ condition: The logit-linked GLMM was used to examine the effect of group and goal position in a reference context, with binary aperture choice as the dependent variable. Group, goal position, and their interaction were included as fixed effects, and a random intercept for each participant was added to account for within-subject dependency, allowing for the assessment of potential baseline differences between groups. A model excluding the interaction term was also estimated, and a likelihood ratio test (ANOVA) showed that the interaction model provided a significantly better fit [χ²(1) = 42.64, p < 0.001]. Multicollinearity among predictors was checked using the variance inflation factor (VIF), which remained below 2, indicating acceptable independence between variables. Model coefficients were exponentiated to obtain odds ratios (ORs) with 95% confidence intervals (Wald method), and p-values were extracted to assess the significance of effects. Post hoc analyses were conducted, including the estimation of slopes for goal position within each group. The slopes quantify the rate of change in participants’ choice probability across goal positions, providing insight into how decisively participants adjust their path selection in response to aperture width and goal location. Pairwise comparisons of estimated marginal trends were performed, with Bonferroni correction applied to account for multiple comparisons.

• ‘Asym’, ‘HumNarrow’ and ‘HumLarge’ conditions: The logit-linked GLMM was used to jointly examine the influence of situational (aperture width) and social (presence or absence of a human interferer) factors, as well as their potential interaction with group status. The binary outcome variable was aperture choice. Fixed effects included goal position, group, condition (‘Asym’, ‘HumNarrow’, and ‘HumLarge’), and all two- and three-way interactions. A random intercept for each participant was included to account for within-subject dependency. A model excluding the interaction term was also estimated, and a likelihood ratio test (ANOVA) showed that the interaction model provided a significantly better fit [χ²(7) = 159.54, p < 0.001]. Multicollinearity among predictors was checked using the variance inflation factor (VIF), which remained below 2, indicating acceptable independence between variables. Post hoc analyses were conducted through pairwise comparisons of estimated marginal trends (slopes), which quantify the rate of change in choice probability across goal positions, providing insight into how decisively participants adjusted their path selection. Pairwise comparisons were performed between groups within each condition and between conditions within each group. Bonferroni correction was applied to adjust for multiple comparisons. For all conditions, switch points were defined as the goal position at which the predicted probability of selecting either aperture was exactly 50%. These points were derived from the fitted logistic models by determining where the logistic function intersected the 0.5 probability threshold, representing the moment of equal likelihood between the two aperture selections.

To address our secondary objectives, analyzing participants’ gait characteristics (speed, clearance distance, shoulder rotation angle and trunk torsion) and to assess the influence of pain perception variables on the decision-making of aperture passed.

• For the ‘Sym’ condition, a repeated measures ANOVA was performed to assess whether participants behaved similarly across groups and goal position. The dependent variable was goal position, and the independent variable was group. The data met the assumptions of normality and homogeneity of variances. When a significant main effect or interaction was detected, pairwise post hoc comparisons using t-tests with Bonferroni correction were conducted to further investigate the differences.

• For the ‘Asym’, ‘HumNarrow’, and ‘HumLarge’ conditions, we performed, for each goal separately, a mixed-design ANOVA to analyze participants’ gait characteristics as dependent variables, with group and conditions as fixed factors. Assumptions of normality and homogeneity of variances were verified and met for all variables. When a significant interaction was found, simple effect ANOVAs with Bonferroni correction were conducted, followed by post hoc pairwise t-tests with Bonferroni correction where appropriate. If only a main effect was significant, pairwise t-tests with Bonferroni correction were performed directly.

• Finally, Spearman correlation tests were performed between the scores obtained on the scales and questionnaires assessing pain-related factors (RMDQ, FABQ, PCS, TSK, PIPS, and VAS) and the percentage of trials in which the large aperture was passed, for each goal and condition. Spearman’s method was chosen due to the non-normal distribution of the percentage of trials in which the large aperture was passed.