Pedestrian Behavior Questionnaire (PBQ)

• PBQ is a 20-item instrument that quantifies risky pedestrian behavior using a 6-point scale across five subscales: Violation, Error, Lapse, Aggressive, and reverse-scaled Positive.
• It serves as a pre-experiment measure to profile baseline pedestrian crossing styles and predict trust formation in naturalistic encounters with Level-4 autonomous vehicles.
• Key findings highlight that higher Error scores are linked to increased trust, while reverse-scaled Positive scores indicate decreased trust, underscoring its role in behavioral profiling.

Searching arXiv for papers mentioning the Pedestrian Behavior Questionnaire and closely related pedestrian behavior profiling work. The Pedestrian Behavior Questionnaire (PBQ) is a 20-item questionnaire measuring risky pedestrian behavior. In the arXiv literature provided here, it appears as a pre-experiment individual-differences instrument used to characterize baseline pedestrian crossing style in a naturalistic study of repeated pedestrian interactions with Level-4 autonomous vehicles (AVs) at an uncontrolled urban intersection (Chang et al., 30 Sep 2025). The available evidence presents PBQ not as a trust scale or an AV-specific attitudinal instrument, but as a behavioral profiling measure whose subscales can help explain heterogeneity in trust formation during real-world pedestrian–AV encounters. Related arXiv work on pedestrian psychometrics does not mention PBQ by name, but it does describe adjacent questionnaire-based constructs—especially personality traits, herding, group coordination, evacuation readiness, route choice, and detour behavior—that overlap conceptually with PBQ-like profiling (Schultz et al., 2012).

1. Instrument definition and measurement scope

In the cited AV study, PBQ is described as a 20-item questionnaire with a 6-point scale, where higher scores represent riskier pedestrian behavior (Chang et al., 30 Sep 2025). The instrument is organized into five subscales: Violation, Error, Lapse, Aggressive, and Positive. The paper states that PBQ was included because the study sought to test whether baseline pedestrian crossing style / risky behavior helps predict how much trust a person develops after repeated exposure to AVs operating safely in real traffic.

This framing assigns PBQ a specific methodological role. It is used to capture a pedestrian’s own behavior tendency in traffic, rather than receptivity to AVs per se or generalized openness to technology. The same study distinguishes PBQ from the Pedestrian Receptivity Questionnaire for Fully AVs (PRQF), which measures AV-specific receptivity, and from the Trust in AVs Scale, which is the primary outcome measure. A plausible implication is that PBQ occupies the behavioral end of the construct space, whereas PRQF and trust scales occupy attitudinal and evaluative roles.

The paper further notes that it relies on prior validation from Deb et al. (2017), where PBQ was developed and validated for pedestrian crossing behavior, and that PBQ is therefore treated as an established external scale rather than being psychometrically revalidated within the robotaxi study itself (Chang et al., 30 Sep 2025).

2. Subscales and score interpretation

The five PBQ subscales reported in the AV study are summarized below.

Subscale	Characterization in the study
Violation	Risky pedestrian behavior subscale
Error	Inadvertent or inattentive crossing mistakes
Lapse	Risky pedestrian behavior subscale
Aggressive	Risky pedestrian behavior subscale
Positive	Safe, socially considerate pedestrian behavior; reverse-scaled

The Error subscale is described as reflecting inadvertent or inattentive crossing mistakes, including walking on cycling paths by mistake, crossing without observing traffic, and misjudging gaps (Chang et al., 30 Sep 2025). The Positive subscale is described as reflecting safe, socially considerate pedestrian behavior, including cautious crossing and politely yielding. The paper emphasizes an important scoring nuance: Positive is reverse-scaled in the original instrument. Accordingly, although the general rule is that higher PBQ scores correspond to riskier pedestrian behavior, the practical direction of interpretation for Positive must be handled carefully.

This reverse-scaling issue is central to any technical reading of PBQ results. A common misconception would be to read the Positive subscale literally, as though a higher score straightforwardly meant “more positive behavior.” The AV study explicitly warns against that interpretation. For this reason, the Positive subscale is best treated as a score whose semantic meaning depends on the coding convention in the underlying instrument rather than on the subscale label alone (Chang et al., 30 Sep 2025).

3. Administration and analytic role in pedestrian–AV trust research

PBQ was administered before the experiment day, together with demographics and the Personal Innovativeness Scale (PIS), in a study where 33 participants repeatedly crossed an uncontrolled urban intersection with frequent Level-4 Robotaxi traffic in a commercial Robotaxi operation zone (Chang et al., 30 Sep 2025). The study was explicitly designed to move beyond simulation-heavy literature by using a naturalistic public-road setting.

Within this design, PBQ functioned as a pre-experiment individual-differences measure, not as a state measure expected to change during the crossing task. The primary dependent variable was trust in AVs, assessed before and after the experiment using a 6-item Trust Scale. The stated logic was that pedestrians differ in their everyday traffic behavior, that those differences may shape how they interpret AV behavior, and that repeated real-world interactions may reinforce or weaken trust depending on that baseline behavior profile.

This use of PBQ is methodologically specific. The study did not present PBQ as a diagnostic of actual trajectory-level motion in the immediate crossing episode. Instead, PBQ served as an explanatory covariate for between-person variation in trust outcomes. This suggests a two-layer model of pedestrian–AV interaction: baseline behavioral disposition measured by PBQ, and interaction-contingent trust measured longitudinally across pre/post exposure.

4. Statistical findings and modeled effects

The AV study fit a mixed linear effects model using SAS Proc Mixed, with repeated measures accounted for using GEE. Predictors included Timing (pre vs. post), PIS, PRQF subscales (Safety, Interaction, Compatibility), PBQ subscales (Violation, Error, Lapse, Aggressive, Positive), and all two-way interactions with Timing. The study reports backward model selection based on BIC (Chang et al., 30 Sep 2025).

Among PBQ variables, only Error and Positive remained significant in the final model. The reported effects were:

• Error: $\beta = 0.09$, SE = 0.04, $F(1,29)=5.99$, $p=.02$
• Positive: $\beta = 0.05$, SE = 0.02, $F(1,29)=5.97$, $p=.02$

The discussion further reports marginal-effect confidence intervals:

• Error: trust increased by 0.09 per 1-unit increase in Error, $95\%\,CI=[0.02,0.20]$
• Positive: trust decreased by 0.05 per 1-unit increase in Positive, $95\%\,CI=[0.01,0.09]$

The paper’s interpretation is that people who are more prone to inadvertent crossing mistakes tended to trust AVs more, and that people who show less “positive” pedestrian behavior also tended to trust AVs more (Chang et al., 30 Sep 2025). Because Positive is reverse-scaled, this is an area where the paper itself urges careful directional interpretation. The same study also reports that trust increased after interaction overall, although the direct main effect of Timing in the final table is not significant: $\beta=-0.56$, SE = 0.43, $F(1,31)=1.74$, $F(1,29)=5.99$0.

The modeled structure is summarized in the source as:

$F(1,29)=5.99$1

where $F(1,29)=5.99$2 indexes participants, $F(1,29)=5.99$3 indexes measurement occasion, $F(1,29)=5.99$4 indexes PRQF subscales, and $F(1,29)=5.99$5 indexes PBQ subscales. In the final reported results, the study does not report a statistically significant PBQ-by-Timing interaction; PBQ effects are described as main effects on trust rather than as effects that differ materially between pre- and post-experiment measurements (Chang et al., 30 Sep 2025).

5. Relation to PRQF, trust scales, and broader pedestrian profiling

The AV study explicitly differentiates three construct classes. PBQ reflects a pedestrian’s own crossing style/risk tendency. PRQF reflects how receptive the pedestrian is to AVs as a road user. The Trust in AVs Scale measures the trust outcome itself. The most prominent PRQF result is a Timing × Interaction effect: participants with higher PRQF Interaction scores reported higher trust both pre- and post-experiment, with the relationship stronger after the experiment. The reported coefficients are 0.69 pre-experiment and 0.82 post-experiment, both with $F(1,29)=5.99$6 (Chang et al., 30 Sep 2025).

In that configuration, PBQ contributes explanatory variance that is complementary to AV-specific receptivity. The paper’s key comparative point is that pedestrian behavior measures like PBQ can matter even when general tech innovativeness does not, since PIS was tested but did not remain among significant effects after model selection. This establishes PBQ as a traffic-behavior trait measure with predictive value distinct from both technology orientation and AV-specific receptivity.

The authors’ interpretation is that PBQ reveals a behavioral mechanism of trust formation through behavioral exposure. Pedestrians with more error-prone behavior may create situations in which AVs must respond safely; because the robotaxis in the study behaved conservatively and predictably, these participants may have received repeated positive feedback from the AVs, which likely increased trust. A similar argument is made for participants scoring higher on the reverse-scaled Positive subscale: if they were less inclined to yield or behave cautiously, they may again have encountered more noticeable safe responses from AVs (Chang et al., 30 Sep 2025). This suggests that trust formation is not only a function of attitude but also of the interaction opportunities induced by the pedestrian’s own behavioral style.

6. Conceptual antecedents, adjacent constructs, and limitations

An important boundary condition is that not all pedestrian psychometric work is PBQ-specific. The paper “Group dynamic behavior and psychometric profiles as substantial driver for pedestrian dynamics” does not mention PBQ, and it does not present a pedestrian behavior questionnaire of its own; instead, it uses the German version of the EPQ-RK and a second questionnaire focusing on anxiety and stress, with reported results based exclusively on the EPQ-RK data (Schultz et al., 2012). Nevertheless, it is directly relevant at the construct level because it studies group dynamic behavior and psychometric profiles as drivers of pedestrian movement at the tactical level.

That work identifies constructs that overlap with what a PBQ-like framework might seek to measure: intra-group coordination, group size effects, leader–follower structure, non-verbal communication, altruistic behavior, yielding precedence for group benefit, herding / following others, synchronized movement, evacuation readiness, route choice, walking speed, speed deviation, and detour index (Schultz et al., 2012). It also examines how Extraversion and Neuroticism relate to readiness to evacuate, actual start time, speed variability, and detour behavior. The authors’ interpretation is that Extraversion may increase reliance on social feedback, potentially delaying action in uncertain situations, while Neuroticism may increase urgency perception, producing earlier evacuation and more direct walking.

This adjacent literature matters because it clarifies what PBQ is not. PBQ, as used in the AV study, is not a full-spectrum instrument for all dimensions of pedestrian movement. It is a behavioral questionnaire oriented toward risky pedestrian behavior, whereas the broader pedestrian-psychology literature also includes personality models, situational stress and anxiety measures, and observational analyses of group coordination. A plausible implication is that PBQ should be understood as one instrument within a larger ecosystem of pedestrian behavior profiling rather than as a comprehensive substitute for tactical, social, and psychometric modeling.

The available arXiv evidence also imposes limits on what can be claimed about PBQ itself. The AV study does not report new psychometric validation such as Cronbach’s alpha, confirmatory factor analysis, factor loadings, fit indices, or measurement invariance. Nor does it provide the full item list or an item-by-item scoring formula. The strongest supported characterization, therefore, is methodologically narrow but clear: PBQ is an established questionnaire for risky pedestrian behavior whose subscales—especially Error and Positive—can be used to explain individual differences in trust toward AVs in naturalistic pedestrian–robotaxi interaction settings (Chang et al., 30 Sep 2025).