Range nudges enhance behavioural adherence to safety and health guidelines – Nature

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Advertisement
Communications Psychology volume 3, Article number: 97 (2025) Cite this article
2788 Accesses
1 Citations
2 Altmetric
Metrics details
Behavioural guidelines sometimes specify only an upper or lower limit, such as speed limits (e.g., ‘60’) or minimum handwashing durations (e.g., ‘20 s’). Limits can produce anchoring effects, biasing judgments toward the values. The distinction between anchoring arising from limits that semantically imply a range (e.g., speed limit ‘60’ implying ‘0–60 km/h’) and those arising from an explicitly stated range (e.g., ‘0–60’) provides insights into how presentation formats affect anchoring. Here, we show that explicitly stating both limits acts as an additional anchor; the Range Nudge—reframing a single limit as a range—reduces non-adherence behaviour compared to presenting only one limit. In online (Study 1: n = 112) and simulated driving tasks (Study 2: n = 31), while the speed limits ‘60’ and ‘0–60’ are logically equivalent, the range led to lower incidences of speeding. Similarly, in handwashing tasks conducted in online (Study 3a: n = 163; Study 3b: n = 484), field (Study 4: n = 38), lab (Study 5a: n = 19), and individual home settings (Study 5b: n = 442), although the limit (‘more than 20 s’) covered a broader time span than the range (‘20–60 s’), the latter prompted a longer handwashing duration. The results suggest that individuals consider limits as recommendations, but the Range Nudge reduces this tendency. Although the findings (seven experiments, total n = 1199) stem from controlled experiments rather than large-scale real-world applications, they offer theoretical insights and practical guidance for using the Range Nudge to enhance adherence to safety and health guidelines.
In everyday life, people are often guided toward particular behaviours by numerical limits—either upper or lower. For example, speed limit signs typically state a maximum speed, such as ‘60 km/h’, instructing drivers not to exceed that value¹. Similarly, hygiene guidelines often set minimum requirements, such as ‘wash your hands for at least 20 s’². While these numerical limits are intended to encourage safe and healthy behaviour, they may also produce unintended effects on judgment and behaviour through the anchoring effect.
The anchoring effect refers to the cognitive bias whereby judgments are influenced by previously encountered numerical values³. Classic demonstrations show that when people are asked to estimate quantities—such as the percentage of African countries in the United Nations—their estimates are significantly higher after being exposed to a high anchor (e.g., 65) than after a low anchor (e.g., 10). This effect is remarkably robust^4,5, even with brief exposure⁶, and is explained by two dominant theories: the insufficient adjustment hypothesis³ and the selective accessibility hypothesis^7,8,9,10. The insufficient adjustment hypothesis suggests that people adjust their estimates from an initially presented anchor. However, owing to cognitive load and limitations in attentional resources, these adjustments tend to be insufficient. Consequently, the final judgement remains biased towards the anchor. The selective accessibility hypothesis posits that presenting an anchor cognitively activates information related to that anchor, thereby influencing judgement. For example, when estimating the price of a product, if a high anchor such as ‘$100’ is provided, individuals are more likely to retrieve information associated with the product being expensive, leading them to estimate a higher price. Importantly, regardless of which theory is adopted, a single anchor can potentially function as a target value.
Anchoring effects are not limited to single values. When tandem anchors—two numbers presented simultaneously—are used, both values can serve as anchors^11,12,13. For example, buyers made higher counteroffers when the seller initially proposed a price range of $100–120 compared to a point offer of $100¹¹. This suggests that both the numerical values can function as anchors.
Previous studies showed that when only an upper limit is presented (e.g., under $150), the value of $150 functions as an anchor, people tend to focus on numbers above it and interpret that value as relatively low^14,15,16. This insight is highly relevant in contexts such as speed regulation. We predict that a sign showing only ‘60 km/h’ serves as an upper anchor, potentially causing people to focus on values above 60 km/h and perceive 60 km/h as relatively slow. Considering the finding obtained in a previous study¹¹, in contrast, presenting a range such as ‘0–60 km/h’ introduces two anchors—both the low and high—which may reduce the perception that 60 km/h is slow and therefore help discourage speeding. Though both ‘less than 60 km/h’ and ‘0–60 km/h’ are logically equivalent, the presence of dual anchors may shift attention more evenly across the range. The comparison between ‘60 km/h’ and ‘0–60 km/h’ allows us to isolate whether the anchoring effect is primarily driven by the semantic implication of a range (i.e., 0–60 km/h implied by the ‘60 km/h’ sign) or by the explicit presentation of the endpoints (i.e., 0 km/h in the ‘0–60 km/h’ sign). If explicit values such as 0 exert an independent influence, they may shift behaviour even when they are not semantically necessary. Moreover, people are generally receptive to range-based information. In negotiations, for example, over half of participants report using price ranges in offers¹⁷. While speed limits are typically framed as upper limits, there are many real-world contexts where lower bounds are also relevant—suggesting that range framing may be both intuitive and effective.
A similar anchoring mechanism may also influence handwashing behaviour. The commonly used guideline ‘more than 20 s’ provides only a lower limit, which may unintentionally cause individuals to treat 20 s as a sufficient and expected duration. This prediction is consistent with findings from prior studies showing that lower-limit framing can induce anchoring. For instance, when credit card statements display a minimum repayment amount, people often anchor on that minimum and repay only the stated amount, resulting in increased long-term interest payments^18,19,20. Unlike speed limits, however, recommended handwashing durations can be contextually adjusted. Research indicates that washing for around 60 s is significantly more effective at removing viruses than washing for only 10 to 30 s²¹. At the same time, excessive handwashing may cause skin irritation²², suggesting that 60 s is a practical upper limit. Thus, reframing the guideline as a numerical range—for example, ‘20–60 s’—introduces an upper anchor in addition to the lower one. This format not only emphasizes the minimum duration but also draws attention to the longer, more effective end of the range. As a result, individuals may be more likely to engage in extended handwashing, improving adherence to hygiene standards.
Prior research has explored several types of anchoring: single and tandem anchors. However, additional research is needed to explore the distinction between anchoring that arises from semantically implied ranges (e.g., ‘60 km/h’ implicitly meaning 0–60) and those that come from explicitly stated ranges (e.g., ‘0–60 km/h’). Thus, we can test whether, even when the two formats are logically equivalent in meaning (e.g., ‘60 km/h’ and ‘0–60 km/h’), the presence of an explicitly stated second value functions as an additional anchor, thereby enhancing compliance with safety guidelines. Similarly, the recommended handwashing duration of ‘more than 20 s’ lacks an explicit upper limit, implicitly suggesting a wide range of acceptable durations (i.e., from 20 s to infinity). In contrast, ‘20–60 s’ clearly defines an upper limit of 60 s, explicitly narrowing the implied range. This comparison allows us to test whether explicitly stating ‘60 s’ functions as an additional anchor, thereby enhancing behavioural adherence to health guidelines.
The present research introduces the concept of a Range Nudge: a behavioural intervention in which a directive that traditionally includes only a single numeric limit (either upper or lower) is reformulated into an explicit numerical range by adding the missing boundary. Nudges are interventions that steer behaviour by modifying choice architecture without restricting options or altering economic incentives²³. The concept of choice architecture refers to how the presentation format often determines which option is selected²⁴. The Range Nudge aims to improve adherence to safety and health protocols—such as speed limits and handwashing—by changing how numerical information is presented.
This approach is not only theoretically motivated but also provides insights into practical applications. For instance, despite the widespread presence of speed limits, there are 88,500 speeding violations per year in Japan²⁵, and 22.7% of U.S. drivers report speeding as their primary risky behaviour²⁶. Similarly, although 50% of South Koreans say they wash their hands for at least 21 s, only 7.6% actually do so²⁷. By explicitly presenting both endpoints (e.g., ‘0–60 km/h’ instead of ‘60 km/h’, or ‘20–60 s’ instead of ‘more than 20 s’), the Range Nudge introduces a second numeric anchor, thereby enhancing behavioural guidance. This may provide a simple yet effective tool for improving adherence to safety and health guidelines.
To evaluate the effectiveness of the Range Nudge, we conducted seven experiments across different modalities: web-based surveys (Studies 1, 3a, and 3b), lab-based simulations (Studies 2 and 5a), and field studies (4 and 5b). Web-based experiments allow for a diverse participant pool, increasing external validity²⁸, while laboratory experiments ensure internal validity by controlling for confounding factors²⁹. Field experiments, with their high environmental realism, enhance the generalizability of findings²⁹. By combining these three methods, this study aims to comprehensively evaluate the effectiveness of the Range Nudge.
Studies 1 and 2 examine whether adding a lower anchor (‘0–60 km/h’) to speed limit signage affects judgment and behaviour compared to the traditional ‘60 km/h.’ Studies 3a–5b shift the focus to hand hygiene, investigating whether adding an upper anchor (‘20–60 s’) increases handwashing duration compared to ‘more than 20 s.’ These experiments employ both between- and within-participants designs to assess whether even small shifts in numeric format can nudge behaviour.
By combining theoretical insights on anchoring with practical interventions grounded in choice architecture, these controlled experiments explore how simple changes in numeric framing can produce improvements in public behaviour.
A web-based questionnaire using a between-participants design was conducted to determine whether different signs (e.g., ‘60 km/h’ vs. ‘0–60 km/h’) significantly affected judgement. The Supplementary Information provides additional details on the experimental materials, procedures, data analysis, and further analyses.
The protocols of all experiments were approved by the Ethics Review Committee for Experimental Research with Human Subjects at the University of Tokyo. All participants provided informed consent prior to participation. All experiments were not pre-registered.
Some estimations are easier than others, depending on the type of target, leading to differences in the size of anchoring effects³⁰. Using publicly available datasets in a meta-analysis of anchoring effects, which considers target type, numerical values and factors like incentives for accurate estimates, the effect size is g = 0.59 when the anchoring task involved ‘speed’³⁰. Hedges’ g³¹ is a statistical measure similar to Cohen’s d³² but it corrects the bias in small sample sizes. However, when the sample size is large (n per group larger than 20), Hedges’ g and Cohen’s d provide nearly identical values³³. Therefore, an effect size of d = 0.59 was chosen, and the required number of participants for Study 1 was calculated. For all experiments, the required sample size was calculated using G*Power³⁴ (version 3.1.9.3), and a sensitivity analysis was conducted with the same software to assess the robustness of the power estimates. A sample size of 98 participants (49 per group) was required to detect an effect size of d = 0.59 (power = 80%, α = .05, two-tailed test, two groups, Wilcoxon–Mann–Whitney test). However, considering the potential for a substantial amount of missing data in web-based experiments, we recruited a larger sample. Thus, 112 Japanese participants, all possessing a driver’s licence (M_age = 48.65 years, SD_age = 10.87; 78 male and 34 female) were recruited through Rakuten Insight (https://insight.rakuten.co.jp/en/), a web-based research company. In all experiments, there were no restrictions on participation based on factors such as age, sex, or place of residence. In all experiments, participants were Japanese, and the demographic data collection focused on their age and sex. Other demographic details such as racial identity, location, ethnicity, nativity or immigration history, socioeconomic status, and other relevant demographics were not measured. After all the tasks were completed, Rakuten Insight paid each participant in points (the authors were not informed of the specific amount), redeemable for purchases on the Rakuten website (https://www.rakuten.com/), an electronic commerce company.
In all experiments, instructions and forms were provided in Japanese. Participants answered three questions:
Question 1: ‘What is the average speed you want to drive at?’
Question 2: ‘What speed do you think you are expected to drive at, on average?’
Question 3: ‘Which of the following sentences most accurately describes your feelings about the driving speed: [the driving speed is high at 60 km/h] or [the driving speed is low at 60 km/h]?’
Study 1 was conducted using Qualtrics (https://www.qualtrics.com), a web-based experimental platform. Participants were randomly assigned to either the ‘60’ sign group (Fig. 1a) or the ‘0–60’ sign group (Fig. 1b).
Panel  (a) presents the sign used for the 60 group. Panel (b) presents the sign used for the 0–60 group.
After being shown the sentence, ‘When you are driving a car on a local road and see this road sign’, participants were presented with their respective signs on their computer screens and answered three questions. Questions 1–3 were presented on separate pages, with Fig. 1a shown to the ‘60’ group and Fig. 1b shown to the ‘0–60’ group. No time limit was imposed for answering any of the questions. Data were collected between January 11, 2021, and February 2, 2021.
Adopting a within-participants design, an experiment was conducted to examine whether differences in signs (e.g., ‘60’ vs. ‘0–60’) significantly affected driving speed, using a driving simulator.
Using publicly available datasets in a meta-analysis of anchoring effects, the effect size is g = 0.59 when the anchoring task involved ‘speed’³⁰. Thus, the effect size d = 0.59 was set in accordance with Study 1, and the number of participants required for this study was calculated. Twenty-six participants were needed to detect an effect size of d = 0.59 (power = 80%, α = .05, two-tailed test, matched two pairs, Wilcoxon signed–rank test). Thus, 31 undergraduate and graduate Japanese students with a driver’s licence (M_age = 21.55 years, SD_age = 1.15; 20 female and 11 male) were recruited. After completing all tasks, each participant received compensation of approximately 1100 Japanese yen (about $7.50).
The velocity observed in driving simulations closely aligns with actual on-road vehicle speeds³⁵ and can reliably forecast on-road vehicle speeds³⁶. Therefore, driving speed was measured in driving simulations at approximately 60 fps.
This study used a driving simulator that allowed for controlled experimental conditions, ensuring a consistent road environment without modifications. A monitor (C27G1/11, 27-inch, AOC Monitors) with a curved screen (radius of curvature: 1,700R) was used to output the virtual reality images. To control the driving simulator, a device (Logicool G29 Driving Force) that replicates the steering wheel and pedals of a real car were used. Using the game engine Unity (Unity 2019.4.3f1 Personal, Unity Technologies, Inc), a city, local road and car in virtual reality were created. This study used the ZENRIN City Asset Series^TM and Standard Assets as material to create the city and driving simulator. The traffic lights were always green. Pedestrians and other vehicles were not included. Fig. 2a shows a bird’s-eye view of the city, Fig. 2b shows a screenshot of the virtual reality environment, and Fig. 2c shows a photograph of the experimental apparatus.
Panel (a) presents a bird’s eye view of the entire length of the driving course. As depicted in panel (a), participants started driving the car from the starting point (i.e., the yellow circle), took the route indicated in blue, and drove to the goal point (i.e., the green circle). In panel (a), demarcations indicate where signs were placed on the road (10 locations in total). Panel (b) presents the image projected onto the screen when participants used the driving simulator. Panels (a, b) exclusively display images for which explicit publication permission has been obtained from the model creators. Panel (c) presents the equipment used in Study 2.
Participants practiced driving on a course without any sign for approximately 20 min to familiarise themselves with the simulator. They completed nine laps on the course, with a one-minute break after each lap. The order in which the signs were presented is described below:
No signs
60 or 0–60 signs
No signs
60 or 0–60 signs (if the 60 signs were presented in the second course, the 0–60 signs were presented in the fourth course, and vice versa)
No signs
0–60 signs
0–60 signs
0–60 signs
No signs
Fourteen participants were shown the ‘60’ sign in the second course (M_age = 21.71 years, SD_age = 0.91; 7 male and 7 female), whereas 14 participants (M_age = 21.50 years, SD_age = 1.40; 11 female and 3 male) were presented the ‘60’ sign in the fourth course. As explained later, data from three participants were excluded; thus, 28 participants were included. The course without signs was the filler. Ten signs were used in each course (Signs 2, 4, 6, 7, and 8). Participants took approximately two minutes to drive around each course. No time limits were imposed during this experiment. Data were collected between January 19, 2021, and April 5, 2021.
This study employed a between-participants design using a web questionnaire to assess whether different instructions (e.g., ‘more than 20 s’ vs. ‘more than 20 s but less than 60 s’) significantly affected judgement.
Using publicly available datasets in a meta-analysis of the anchoring effect, the effect size of g = 0.94 was observed when the type of anchoring task is ‘duration’³⁰. Thus, this study set the effect size d = 0.94 and calculated the number of participants required for the experiment. Forty participants were required to detect an effect size of d = 0.94 (power = 80%, α = .05, two-tailed test, two groups, Wilcoxon–Mann–Whitney test). Considering the likelihood of a high amount of missing data in web-based experiments, the number of participants recruited exceeded that determined by the sample size analysis. Thus, 163 Japanese participants (M_age = 45.74 years, SD_age = 12.83; 93 male and 70 female) were recruited from Rakuten Insight. Participants who took part in Study 1 were excluded from participating in this experiment. After all the tasks were completed, Rakuten Insight paid each participant in points (the authors were not informed of the specific amount), redeemable for purchases on the Rakuten website.
Participants answered two questions:
Question 1: ‘How many seconds do you want to wash your hands for?’
Question 2: ‘How many seconds do you think you are expected to wash your hands for?’
An experiment was conducted using Qualtrics.
Participants were randomly assigned to one of two groups—lower-limit and defined-range groups—and presented with the following instruction: ‘Imagine that your friend is moving to a different house, and you are helping them move some furniture. Afterwards, you realise that you touched the furniture with your bare hands, and now you want to wash your hands with soap. Later, you read the following sentence: It is recommended that you wash your hands for more than 20 s’ (for the lower-limit group). For the defined-range group, the instruction ‘more than 20 s’ was modified to ‘more than 20 s but less than 60 s’. After the above instruction, they answered the two questions. No time limits were imposed during this experiment. Data were collected between July 1, 2021, and July 27, 2021.
The sensitivity analysis indicated that study 3a’s sample size allowed detection of effect sizes of d = 0.45 or greater. However, since the effect size observed for Question 1 (d = 0.28) was below this threshold, Study 3a may have lacked sufficient power to reliably detect such a small effect. To ensure a more precise estimate of the effect and to confirm the findings, Study 3b was conducted with an increased sample size. The number of participants was determined based on the effect size observed in Study 3a for Question 1. Upon completing all tasks, each participant received a reward of approximately 20 Japanese yen (about $0.15) for their participation.
A sample size of 422 participants (211 per group) was required to detect an effect size of d = 0.28, which was observed for Question (power = 80%, α = 0.05, two-tailed test, two groups, Wilcoxon–Mann–Whitney test). Thus, 484 participants (M_age = 41.53 years, SD_age = 10.77; 245 female, 234 male, and 5 who preferred not to disclose their sex) were recruited through Crowd Works (https://crowdworks.co.jp/en/), a web-based platform for recruiting research participants.
The same measures as in Study 3a were used.
The same materials as in Study 3a were used.
The procedure was identical to that of Study 3a. Data were collected between January 4, 2025, and January 12, 2025.
A between-participants design was adopted in this behavioural experiment to assess whether differences in instructions (e.g., ‘more than 20 s’ or ‘more than 20 s but less than 60 s’) significantly affected handwashing duration. We assessed whether the different instructions influenced handwashing duration even when different groups of participants washed their hands simultaneously and when others were waiting their turn. Students from the Department of Psychological and Sociological Studies at a different university from that of the authors’ affiliation, who were enrolled in a course on Experimental Psychology, volunteered to participate in the experiment after their lectures.
Study 4 was a field study, and it was unclear how many participants would take part before the experiment began. Consequently, the sample size required for this study was not calculated in advance. Instead, after the experiment was conducted, a sensitivity analysis was performed to ensure that the sample size used was sufficient to detect the observed effect size with adequate statistical power. The results of the sensitivity analysis are reported in the Results section. In total, 38 participants (M_age = 19.76 years, SD_age = 0.91; 27 female and 10 male) were recruited.
Participants’ faces and their handwashing procedures were videotaped using a camera positioned at the side of the handwashing area with their consent. To prevent experimenter bias, a collaborator unaware of this study’s aim recorded the video. The collaborator was asked to measure each participant’s handwashing durations and the number of participants who were standing in line behind them while they washed their hands.
Four water faucets and soaps were provided, allowing a maximum of four participants to wash their hands simultaneously.
All participants gathered in the gymnasium and picked up items to necessitate handwashing (e.g., measuring tapes and sticky notes that were scattered on the floor), then were randomly assigned to the lower-limit or defined-range group. They received the following instruction (for the lower-limit group): ‘To prevent infection, please wash your hands. The washroom is located at the end of the stairs in the gymnasium. It is recommended that you wash your hands for more than 20 s. The target duration for handwashing is only a guideline, and you should decide how long to wash your hands for’. For the defined-range group, the instruction ‘more than 20 s’ was modified to ‘more than 20 s but less than 60 s’. Then, those who were prepared to wash their hands did so at the water faucet using hand soap. Participants in both groups washed their hands together. When all four faucets were in use, participants who had not washed their hands waited in line. No time limits were imposed during this experiment. The experiment was conducted on July 2, 2021, during the COVID-19 pandemic.
Following a within-participants design, a behavioural experiment was individually conducted to assess whether differences in instructions (e.g., ‘more than 20 s’ vs. ‘more than 20 s but less than 60 s’) significantly affected handwashing duration.
Using publicly available datasets in a meta-analysis of the anchoring effect, the effect size of g = 0.94 was observed when the type of anchoring task is ‘duration’³⁰. The effect size d = 0.94 was set and the number of participants required for this study was calculated. Twelve participants were required to detect an effect size of d = 0. 94 (power = 80%, α = .05, two-tailed test, matched two pairs, Wilcoxon signed–rank test). Thus, 19 undergraduate and graduate students (M_age = 21.32 years, SD_age = 1.83; 10 female and 9 male) were recruited. Participants who took part in Studies 2 and 4 were excluded from participating in this experiment. Upon completion of all tasks, each participant was paid approximately 1000 Japanese yen (about $7) for their participation.
Handwashing procedures were videotaped using a camera positioned at the side of the handwashing area with participants’ consent. To prevent experimenter bias, a collaborator unaware of this study’s aim recorded the video. The collaborator was asked to measure each participant’s handwashing durations.
Participants washed their hands with hand soap in the handwashing area of the laboratory.
The experiment was conducted by randomly determining whether to first present the ‘more than 20 s’ notation to participants or to present it to them later. Of the final 18 participants, 9 were first shown the ‘more than 20 s’ notation (M_age = 21.56 years, SD_age = 1.59; 6 male and 3 female), and 9 (M_age = 21.11 years, SD_age = 2.21; 6 female and 3 male) were presented the notation later. As explained later, data from 1 participant were excluded; thus, 18 participants were included.
This experiment was conducted on an individual basis. The following instructions were presented to participants. Instruction 1 (for the lower-limit condition): ‘To avoid contracting COVID-19, you are required to wash your hands with soap before the experimental task. Please use the handwashing area in the laboratory. It is recommended that you wash your hands for more than 20 s to thoroughly remove viruses and dirt. The target duration for handwashing is only a guideline, and you should decide how long to wash your hands’. For the defined-range condition, the instruction ‘more than 20 s’ was modified to ‘more than 20 s but less than 60 s’. After participants washed their hands using hand soap, they performed an unrelated experimental task for approximately 40 min. Subsequently, instruction 2 was presented to participants. Instruction 2 was identical to instruction 1, with the following variation: ‘you are required to wash your hands with soap after the experimental task’ was substituted for ‘you are required to wash your hands with soap before the experimental task’. When ‘more than 20 s’ was presented in instruction 1, ‘more than 20 s but less than 60 s’ was presented in the second instruction and vice versa. No time limits were imposed during this experiment. Data were collected between November 18, 2021, and December 9, 2021, during the COVID-19 pandemic.
Sensitivity analysis indicated that study 5a’s sample size allowed detection of effect sizes of d_z = 0.70 or greater. The effect size d_z represents standardized difference scores³⁷. However, since the effect size (d_z = 0.30) observed was below this threshold, study 5a may have lacked sufficient power to detect such a small effect reliably. To ensure a more precise estimate of the effect and to confirm the findings, Study 5b was conducted with an increased sample size. To recruit a larger number of participants, we modified the procedure from Study 5a. Instead of conducting the experiment in person, participants recorded a video of themselves washing their hands and submitted it. Since participants washed their hands at their preferred times, it would have been difficult to control the time interval between the first and second handwashing sessions in a within-participants design with different instructional conditions. Therefore, we adopted a between-participants design instead.
A total of 352 participants were required to detect an effect size of d = 0.30, which was observed in Study 5a (power = 80%, α = .05, two-tailed test, two groups, t-test). Thus, 442 participants (M_age = 42.55 years, SD_age = 12.01; 229 male, 207 female, and 6 who preferred not to disclose their sex) were recruited through Lancers (https://www.lancers.co.jp/en/), Crowd Works, and Yahoo! Crowdsourcing (https://crowdsourcing.yahoo.co.jp/), a web-based platforms for recruiting research participants. Upon completing all tasks, each participant received a reward of approximately 100 Japanese yen (about $0.7) for their participation.
The participants recorded a video of themselves washing their hands at home. To prevent experimenter bias, a collaborator who was unaware of this study’s aim was asked to measure each participant’s handwashing duration.
A single faucet was used, and each participant recorded their handwashing with their own smartphone or other device. Study 5b was conducted using Qualtrics.
Participants were randomly assigned to one of two groups—lower-limit and defined-range groups—and presented with the following instruction: ‘Currently, influenza is spreading. Therefore, we are conducting a study on how people wash their hands. When washing your hands, please do so alone without any other individuals present. It is recommended that you wash your hands for more than 20 s. The recommended handwashing duration is only a guideline, so please decide the duration based on your own judgement’. For the defined-range group, the instruction ‘more than 20 s’ was modified to ‘more than 20 s but less than 60 s’. Participants recorded a video of themselves washing their hands using a smartphone or other device and then sent the video to the researchers. No time limits were imposed during this experiment. Data were collected between December 26, 2024, and January 23, 2025, during the influenza pandemic.
In all experiments, data analyses were conducted in R³⁸ (version 4.2.2). Data from two participants in the 60 group who did not provide numerical values were excluded from the analysis. Thus, the data from 55 participants in the 60 group (M_age = 48.42 years, SD_age = 10.74; 34 male and 11 female) and 55 participants in the 0–60 group (M_age = 48.82 years, SD_age =10.94; 42 male and 13 female) were analysed.
Figure 3a shows the results for Question 1: ‘What is the average speed you want to drive at?’ The Shapiro–Wilk test indicated non-normal distribution of the data (w = 0.85, p < 0.001). Levene’s F-test for equal variances indicated no significant difference between groups (F [54, 54] = 1.11, p = 0.697), confirming the homogeneity of variances assumption. The Wilcoxon–Mann–Whitney test revealed that the speed was significantly lower in the 0–60 group (mean = 57.98 km/h, median = 60 km/h [interquartile range—IQR—55 to 60 km/h], mean rank = 43.92) than in the 60 group (mean = 62.56 km/h, median = 60 km/h [IQR, 60 to 65 km/h], mean rank = 67.08, p < 0.001, Z = 4.24, r = 0.40, 95% CI [0.23, 0.55], d = 0.72, 95% CI [0.33, 1.11]). Although the analysis was conducted using the Wilcoxon–Mann–Whitney test, mean values, and effect size d, which were used for sample size determination, are also reported.
The y-axis in panel (a) demonstrates the speed that participants wanted to drive at. The y-axis in panel (b) presents the speed that participants believed they were expected to drive at. In panels (a, b), the x-axis indicates the type of sign used. The centre line represents the median, while the edges of the box mark the 25th percentile (lower quartile) and the 75th percentile (upper quartile). The whiskers extend to 1.5 times the IQR beyond the quartiles, and any data points outside this range are classified as outliers. ***p < 0.001.
Figure 3b shows the results for Question 2: ‘What speed do you think you are expected to drive at, on average?’ The Shapiro–Wilk test revealed that the data were not normally distributed (w = 0.82, p < 0.001). Levene’s F-test for equal variances indicated no significant difference between groups (F [54, 54] = 1.32, p = 0.311), confirming the homogeneity of variances assumption. The Wilcoxon–Mann–Whitney test revealed that the speed was significantly lower in the 0–60 group (mean = 53.90 km/h, median = 55 km/h [IQR, 50 to 60 km/h], mean rank = 45.05) than in the 60 group (mean = 59 km/h, median = 60 km/h [IQR, 60 to 60 km/h], mean rank = 65.95, p < 0.001, Z = 3.83, r = 0.36, 95% CI [0.19, 0.52], d = 0.79, 95% CI [0.40, 1.19]).
For Question 3: ‘Which of the following sentences most accurately describes your feelings about the driving speed: [the driving speed is high at 60 km/h] or [the driving speed is low at 60 km/h]?’ 35/55 (approximately 64%) of respondents answered that [the driving speed is low at 60 km/h] in the 60 group. However, in the 0–60 group, 34/55 participants (approximately 62%) answered that [the driving speed is high at 60 km/h]. Pearson’s chi-square test revealed a significant difference between the two groups in the selection ratio (χ²[1] = 6.15, p = 0.013, φ = 0.26, 95% CI [0.07, 0.42]).
A sensitivity analysis determined the minimum detectable effect size, given this study’s sample size and statistical parameters. The analysis indicated that this study could detect an effect size of d = 0.55 (110 participants [55 per group], power = 80%, α = 0.05, two-tailed test, two groups, Wilcoxon–Mann–Whitney test). The observed effect sizes for Question 1 (d = 0.72) and Question 2 (d = 0.79) exceeded this threshold, suggesting that this study had sufficient power to detect meaningful differences between groups.
In the Supplementary Information, a linear mixed model (LMM) analysis was conducted to examine whether the Range Nudge effect remained significant after controlling for random factors, including age and sex.
Three participants’ data were excluded from the analysis owing to 1) incomplete data because of a faulty connection in the experimental apparatus, 2) one participant self-reporting that they intentionally increased their speed in the second half of the experiment and 3) excessive speed causing the virtual vehicle to crash. Therefore, the data for 28 participants (M_age = 21.61 years, SD_age = 1.17; 18 female and 10 male) were analysed in this study.
We obtained and analysed velocity data at each coordinate along the straight sections of the course. The driving speed was measured at approximately 60 fps. For each road segment, we retained a single observation for each unique integer coordinate (X or Z) based on the vehicle’s coordinate system, and removed all other duplicate-coordinate data during the straight sections. A violin plot based on the raw data prior to duplicate removal is provided in the Supplementary Information (see Supplementary Fig. 5). Consequently, each participant contributed 2285 data points during the straight sections of each course. Fig. 4 illustrates the velocities categorized by the speed limit signs. While the coordinates along the straight sections were identical for all participants, data from the turns varied due to differences in how participants navigated the turns. Specifically, variations in turning radius affected the total travel distance, leading to differences in trajectory. Therefore, data points corresponding to the corner sections were removed. Consequently, gaps appear in the data at turning points in the figure.
The y-axis presents mean velocity. The x-axis indicates the number of coordinate changes. The blue line graph shows the 0–60 condition and the orange line graph shows the 60 condition. As each course yielded 2,285 data points, a total of 63,980 data points were obtained for each group with 28 participants. Error bars represent the ±1 standard deviation. The position labelled ‘Sign’ indicates where a sign is present.
We extracted data only for cases in which the speed limit of 60 km/h was exceeded (i.e., 61 km/h and above) and compared the data for each group. For the 60 km/h signs, 11,505 data points were extracted, and for 0–60 km/h signs, 8,712 data points were extracted. These numbers represent the total data points where all participants exceeded the respective speed limits for each condition. Normality of the speed variable was assessed with the Shapiro–Wilk test on a random subsample of N = 5000 observations (the R implementation of Shapiro–Wilk is limited to n ≤ 5000) and revealed a significant departure from normality (w = 0.69, p < 0.001). Homogeneity of variances across the two conditions was evaluated with the Levene’s F-test for equal variances indicated a significant difference between conditions (F [8711, 11504] = 0.40, p  < 0.001), supporting the violation of the homogeneity of variances assumption. The Wilcoxon signed-rank test revealed that the speed was significantly lower in the 0–60 km/h group (mean = 67.12 km/h, median = 64.62 km/h [IQR, 62.14 to 69.26 km/h], mean rank = 9071.13) than in the 60 km/h group (mean = 70.38 km/h, median = 66.03 km/h [IQR, 63.17 to 72.81 km/h], mean rank = 10894.91, p = 0.031, Z = 2.14, r = 0.42, 95% CI [0.03, 0.68], d = 0.40, 95% CI [0.08, 0.73]).
A sensitivity analysis indicated that this study could detect an effect size of d = 0.04 (a sample of 8712 and 11,505 data points, power = 80%, α = 0.05, two-tailed test, two groups, Wilcoxon–Mann–Whitney test). These data points refer specifically to cases where participants exceeded the speed limit of 60 km/h (i.e., 61 km/h and above), as extracted for each group (8712 for the 0–60 signs and 11,505 for the 60 signs). The observed effect size (d = 0.40) suggests that this study had sufficient power to detect meaningful differences between groups.
We examined whether participants reduced their speed immediately after passing a sign. We extracted speed data from 50 coordinate points behind the sign and calculated the difference between these speeds and the speed recorded at the sign’s location (14,000 data points [50 coordinate points(times)10 signs(times)28 participants] per condition). For example, if a participant’s speed when passing a sign was 60 km/h and their speed at the next coordinate was 59 km/h, the recorded difference would be −1 km/h. The choice of 50 coordinate points was made because the shortest distance between signs was approximately 150 coordinates, meaning that 50 coordinates represented one-third of this distance. We examined the effect of sign presentation on speed adjustments following sign appearance. The Wilcoxon signed-rank test revealed no significant difference between the 0–60 km/h group (mean = 0.31 km/h, median = 0.12 km/h [IQR, −0.13 to 2.34 km/h], mean rank = 14174.85) and the 60 km/h group (mean = 0.34 km/h, median = 0.11 km/h [IQR, −0.48 to 1.92 km/h], mean rank = 13826.15, p = 0.493, Z = 0.71, r = 0.13, 95% CI [−0.22, 0.52], d = 0.01, 95% CI [−0.55, 0.58]).
In the Supplementary Information, generalised linear mixed model (GLMM) and linear mixed model (LMM) analyses were conducted to examine whether the Range Nudge effect remained significant after controlling for random factors, including participants (individual differences), age, stimulus order, sex, and driving experience (i.e., driving licence tenure, weekly driving frequency, and total driving hours). The following additional analyses, which are detailed in the Supplementary Information, explored various aspects of drivers’ responses to the signs. An analysis of vehicle speeds, including the average speed, maximum speed, time spent exceeding the limit, and distance travelled above the limit (as shown in Supplementary Fig. 4), consistently demonstrated the impact of the Range Nudge. An examination of driving speed distributions (as shown in Supplementary Fig. 5) after exposure to the ‘0–60 km/h’ sign showed no instances of extremely low speeds. Further analysis of drivers’ awareness (as presented in Supplementary Table 2) revealed that fewer participants aimed to reach or exceed 60 km/h with the ‘0–60 km/h’ sign compared to the ‘60 km/h’ sign.
This study analysed 82 participants’ data in the lower-limit group (M_age = 45.09 years, SD_age = 13.46; 42 male and 40 female) and 81 participants’ data in the defined-range group (M_age = 46.41 years, SD_age = 12.21; 51 male and 30 female).
Figure 5a illustrates the answers to Question 1: ‘How many seconds do you want to wash your hands for?’ The Shapiro–Wilk test revealed that the data were not normally distributed (w = 0.74, p < 0.001). Levene’s F-test for equal variances indicated no significant difference between groups (F [81, 80] = 1.09, p = 0.706), confirming the homogeneity of variances assumption. The Wilcoxon–Mann–Whitney test revealed that the duration for which participants wanted to wash their hands was significantly longer for the defined-range group (mean = 27.35 s, median = 30 s [IQR, 20 to 30 s], mean rank = 90.51) than for the lower-limit group (mean = 23 s, median = 20 s [IQR, 20 to 30 s], mean rank = 73.59, p = 0.017, Z = 2.38, r = 0.19, 95% CI [0.03, 0.33], d = 0.28, 95% CI [−0.03, 0.59]).
The y-axis in panel (a) presents the duration for which participants wanted to wash their hands. The y-axis in panel (b) presents the duration for which participants believed they were expected to wash their hands. In panels (a, b), the x-axis indicates the differences in the values presented in the instructions. ‘20 s’ refers to the group labelled ‘more than 20 s’, and ‘20–60 s’ refers to the group labelled ‘more than 20 s but less than 60 s’. The centre line represents the median, while the edges of the box mark the 25th percentile (lower quartile) and the 75th percentile (upper quartile). The whiskers extend to 1.5 times the IQR beyond the quartiles, and any data points outside this range are classified as outliers. *p < 0.05, ***p < 0.001.
Figure 5b presents the results of Question 2: ‘How many seconds do you think you are expected to wash your hands for?’ The Shapiro–Wilk test revealed that the data were not normally distributed (w = 0.83, p < 0.001). Levene’s F-test for equal variances indicated no significant difference between groups (F [81, 80] = 0.69, p = 0.073), confirming the homogeneity of variances assumption. The Wilcoxon–Mann–Whitney test revealed that the duration that participants believed they were expected to wash their hands for was much longer for the defined-range group (mean = 37.11 s, median = 30 s [IQR, 30 to 50 s], mean rank = 106.03) than for the lower-limit group (mean = 23.18 s, median = 20 s [IQR, 20 to 20 s], mean rank = 58.26, p < 0.001, Z = 6.73, r = 0.53, 95% CI [0.41, 0.63], d = 1.03, 95% CI [0.70, 1.36]).
A sensitivity analysis indicated that this study could detect an effect size of d = 0.45 (82 and 81 participants, power = 80%, α = 0.05, two-tailed test, two groups, Wilcoxon–Mann–Whitney test). The observed effect sizes for Question 1 (d = 0.28) and Question 2 (d = 1.03) suggest that this study had adequate power to detect the observed effect in Question 2 but may have had limited statistical power to detect the effect in Question 1.
In the Supplementary Information, linear model (LM) and linear mixed model (LMM) analyses were conducted to examine whether the Range Nudge effect remained significant after controlling for random factors, including age and sex.
This study analysed 233 participants’ data in the lower-limit group (M_age = 40.88 years, SD_age = 10.60; 117 female, 113 male, and 3 who preferred not to disclose their sex) and 251 participants’ data in the defined-range group (M_age = 42.13 years, SD_age = 10.91; 128 female, 121 male, and 2 who preferred not to disclose their sex).
Figure 6a illustrates the answers to Question 1: ‘How many seconds do you want to wash your hands for?’ The Shapiro–Wilk test revealed that the data were not normally distributed (w = 0.90, p < 0.001). Levene’s F-test for equal variances indicated no significant difference between groups (F [232, 250] = 0.52, p < 0.001), supporting the violation of the homogeneity of variances assumption. The Wilcoxon–Mann–Whitney test revealed that the duration for which participants wanted to wash their hands was significantly longer for the defined-range group (mean = 29.89 s, median = 30 s [IQR, 20 to 37.5 s], mean rank = 285.74) than the lower-limit group (mean = 21.76 s, median = 20 s [IQR, 15 to 30 s], mean rank = 195.92, p < 0.001, Z = 7.22, r = 0.33, 95% CI [0.25, 0.41], d = 0.67, 95% CI [0.49, 0.85]).
The y-axis in panel (a) presents the duration for which participants wanted to wash their hands. The y-axis in panel (b) presents the duration for which participants believed they were expected to wash their hands. In panels (a, b), the x-axis indicates the differences in the values presented in the instructions. ‘20 s’ refers to the group labelled ‘more than 20 s’, and ‘20–60 s’ refers to the group labelled ‘more than 20 s but less than 60 s’. The centre line represents the median, while the edges of the box mark the 25th percentile (lower quartile) and the 75th percentile (upper quartile). The whiskers extend to 1.5 times the IQR beyond the quartiles, and any data points outside this range are classified as outliers. ***p < 0.001.
Figure 6b presents the results of Question 2: ‘How many seconds do you think you are expected to wash your hands for?’ The Shapiro–Wilk test revealed that the data were not normally distributed (w = 0.84, p < 0.001). Levene’s F-test for equal variances indicated a significant difference between groups (F [232, 250] = 0.26, p < 0.001), supporting the violation of the homogeneity of variances assumption. The Wilcoxon–Mann–Whitney test revealed that the duration that participants believed they were expected to wash their hands for was much longer for the defined-range group (mean = 37.69 s, median = 40 s [IQR, 30 to 50 s], mean rank = 316.82) than for the lower-limit group (mean = 23.02 s, median = 20 s [IQR, 20 to 25 s], mean rank = 162.44, p < 0.001, Z = 12.64, r = 0.58, 95% CI [0.51, 0.63], d = 1.30, 95% CI [1.10, 1.49]).
A sensitivity analysis indicated that this study could detect an effect size of d = 0.26 (233 and 251 participants, power = 80%, α = 0.05, two-tailed test, two groups, Wilcoxon–Mann–Whitney test). The observed effect sizes for Question 1 (d = 0.67) and Question 2 (d = 1.30) suggest that this study had adequate power to detect the observed effect.
In the Supplementary Information, a linear mixed model (LMM) analysis was conducted to examine whether the Range Nudge effect remained significant after controlling for random factors, including age and sex.
The lower-limit group comprised 18 participants (M_age = 19.72 years, SD_age = 0.96; 10 female and 8 male), while the defined-range group comprised 20 participants (M_age = 19.80 years, SD_age = 0.89; 17 female and 3 male).
The handwashing durations are presented in Fig. 7. The Shapiro–Wilk test revealed that handwashing duration was normally distributed (w = 0.96, p = 0.211). Levene’s F-test for equal variances indicated a significant difference between groups (F [17, 19] = 0.25, p = 0.006), supporting the violation of the homogeneity of variances assumption. Welch’s two-sample t-test—which adjusts the degrees of freedom to account for variance heterogeneity—revealed that handwashing duration was significantly longer for the defined-range group (M = 37.56 s [95% CI, 32.38 to 42.74 s]) than for the lower-limit group (M = 28.14 s [95% CI, 25.38 to 30.90 s], t[28.58] = 3.36, p = 0.002, d = 1.06, 95% CI [0.35, 1.76]).
The y-axis presents participants’ handwashing duration. The x-axis indicates the differences in the values presented in the instructions. ‘20 s’ refers to the group labelled ‘more than 20 s’, and ‘20–60 s’ refers to the group labelled ‘more than 20 s but less than 60 s’. The centre line represents the median, while the edges of the box mark the 25th percentile (lower quartile) and the 75th percentile (upper quartile). The whiskers extend to 1.5 times the IQR beyond the quartiles, and any data points outside this range are classified as outliers. **p < 0.01.
A sensitivity analysis indicated that this study could detect an effect size of d = 0.94 (18 and 20 participants, power = 80%, α = 0.05, two-tailed test, two groups, t-test). The observed effect sizes (d = 1.06) suggest that this study had adequate power to detect the observed effect.
In the Supplementary Information, a linear mixed model (LMM) analysis was conducted to examine whether the Range Nudge effect remained significant after controlling for random factors, including the number of individuals who were waiting to wash their hands, age, and sex.
The data from one participant who performed another task (i.e., gargling) while washing their hands were excluded from the analysis. Therefore, the data from 18 participants (M_age = 21.33, SD_age = 1.88; 9 male and 9 female) were analysed.
The handwashing durations are presented in Fig. 8. The Shapiro–Wilk test revealed that handwashing duration was normally distributed (w = 0.98, p = 0.578). Levene’s F-test for equal variances indicated no significant difference between conditions (F [17, 17] = 1.15, p = 0.775), confirming the homogeneity of variances assumption. A paired-samples t-test revealed that the duration was significantly greater for the defined-range condition (M = 42.89 s [95% CI, 35.79 to 50.00 s]) than for the lower-limit condition (M = 38.48 s [95% CI, 30.86 to 46.10 s], t[17] = 2.49, p = 0.024, d_z = 0.30, 95% CI [0.05, 0.54]).
The y-axis presents participants’ handwashing duration. The x-axis indicates the differences in the values presented in the instructions. The sections labelled ‘20 s’ represent the combined data from participants who were first presented with the ‘more than 20 s’ notation and those who were presented with it later. Similarly, the sections labelled ‘20–60 s’ represent the combined data from participants who were first presented with the ‘more than 20 s but less than 60 s’ notation and those who were presented with it later. The centre line represents the median, while the edges of the box mark the 25th percentile (lower quartile) and the 75th percentile (upper quartile). The whiskers extend to 1.5 times the IQR beyond the quartiles, and any data points outside this range are classified as outliers. *p < 0.05.
A sensitivity analysis indicated that this study could detect an effect size of d_z = 0.70 (18 participants, power = 80%, α = 0.05, two-tailed test, two conditions, paired-samples t-test). However, the observed effect size (d_z = 0.30) suggests that this study did not have adequate power to detect the observed effect.
In the Supplementary Information, a linear mixed model (LMM) analysis was conducted to examine whether the Range Nudge effect remained significant after controlling for random factors, including participants (i.e., individual differences), age, stimulus order and sex.
Three participants who did not record their handwashing process until completion and one participant who did not respond affirmatively to the comprehension check question were excluded from the analysis. A total of 233 participants were assigned to the lower-limit group (M_age = 43.24 years, SD_age = 12.20; 130 male, 100 female, and 3 who preferred not to disclose their sex) and 205 participants were assigned to the defined-range group (M_age = 41.65 years, SD_age = 11.76; 106 female, 96 male, and 3 who preferred not to disclose their sex).
The handwashing durations are presented in Fig. 9. The Shapiro–Wilk test revealed that handwashing duration was not normally distributed (w = 0.97, p < 0.001). Levene’s F-test for equal variances indicated a significant difference between groups (F [232, 204] = 0.74, p = 0.025), supporting the violation of the homogeneity of variances assumption. The Wilcoxon–Mann–Whitney test revealed that handwashing duration was significantly longer for the defined-range group (mean = 30.58 s, median = 30 s [IQR, 20 to 39 s], mean rank = 239.13) than for the lower-limit group (mean = 26.67 s, median = 25 s [IQR, 18 to 33 s], mean rank = 202.23, p = 0.002, Z = 3.04, r = 0.15, 95% CI [0.05, 0.24], d = 0.31, 95% CI [0.12, 0.50]).
The y-axis presents participants’ handwashing duration. The x-axis indicates the differences in the values presented in the instructions. ‘20 s’ refers to the group labelled ‘more than 20 s’, and ‘20–60 s’ refers to the group labelled ‘more than 20 s but less than 60 s’. The centre line represents the median, while the edges of the box mark the 25th percentile (lower quartile) and the 75th percentile (upper quartile). The whiskers extend to 1.5 times the IQR beyond the quartiles, and any data points outside this range are classified as outliers. **p < 0.01.
A sensitivity analysis indicated that this study could detect an effect size of d = 0.28 (233 and 205 participants, power = 80%, α = 0.05, two-tailed test, two conditions, Wilcoxon-Mann-Whitney test). The observed effect size (d = 0.31) suggests that this study had adequate power to detect the observed effect.
In the Supplementary Information, a linear mixed model (LMM) analysis was conducted to examine whether the Range Nudge effect remained significant after controlling for random factors, including age and sex.
Limits are set to encourage adherence to safety and health protocols. However, this controlled experiment identified situations in which specific limits can unintentionally lead to non-adherence, such as insufficient reductions in driving speed or inadequate increases in handwashing duration. When numerical ranges are presented, both endpoints can serve as anchors^11,12,13. However, additional research was needed to examine whether numerical notation or the semantic range of a given value has a stronger influence on anchoring effects, particularly in cases where upper limits (e.g., driving at or below 60 km/h) and defined ranges (e.g., driving within 0–60 km/h) are logically equivalent. Further, additional research was needed to explore cases in which the semantic range of the lower limit (e.g., washing hands for at least 20 s, meaning from 20 s to infinity) is broader than that of a defined range (e.g., washing hands for 20 to 60 s). Studies 1 and 2 used two logically equivalent pieces of information: the ‘60’ and ‘0–60’ signs. However, participants’ preferred and actual driving speeds were significantly slower when using the defined range than with the upper limit. These results suggest that the presence of the number 0 can influence judgment and behaviour, even when the information presented is logically equivalent. In Studies 3–5, the semantic range of the lower limit (from 20 s to infinity) was wider than that of the defined range (from 20 s to 60 s). However, both desired and actual handwashing durations were significantly longer when using the defined range than with the lower limit. These findings highlight that in the context of the anchoring effect, the influence of the defined range (indicated by numerical values) on judgment and behaviour is stronger than that of the semantic range (indicated by upper/lower limits alone).
Below we discuss how our findings relate to the two underlying mechanisms of the anchoring effects.
Insufficient adjustment³: When a driver intending to travel at 100 km/h encounters a speed limit sign of ‘60 km/h’, the adjustment process starts from ‘60 km/h’ towards 100 km/h. Consequently, drivers are likely to exceed the limit, often choosing a speed somewhere between 60 km/h and 100 km/h. On the other hand, even drivers who initially intend to travel below 60 km/h may treat a 60 km/h speed limit as an anchor. As a result, they may accelerate towards that target and adopt speeds close to the limit. By contrast, when a driver encounters a speed limit sign of ‘0–60 km/h’, the adjustment process starts at ‘0’, proceeds towards the speed limit of 60 km/h, and then continues towards the driver’s intended speed (e.g., 100 km/h). As a result, drivers tend to adjust insufficiently towards 100 km/h compared to when the adjustment starts from 60 km/h. These results demonstrate that even when the information presented is logically equivalent (e.g., ‘less than 60 km/h’ and ‘0–60 km/h’), the explicit presentation of an additional limit (i.e., 0 in ‘0–60 km/h’) exerts a stronger influence on the adjustment process than the semantic range implied by a single limit (i.e., ‘60 km/h’).
Selective accessibility^7,8,9,10: Exposure to a speed limit sign of ‘60 km/h’ may increase the accessibility of information related to speeding, potentially encouraging drivers to adopt speeds close to the limit. In contrast, exposure to a sign ‘0–60 km/h’—through the explicit presentation of the lower anchor of 0—may selectively activate information related to the lower bound, such as ‘very low speeds’ or ‘safe driving’. Consequently, variations in accessible information due to range notation can explain differences in chosen speeds. This finding reveals that even when the information presented is logically equivalent (e.g., ‘less than 60 km/h’ and ‘0–60 km/h’), the signs ‘60 km/h’ and ‘0–60 km/h’ activate different sets of information.
Apart from tandem anchoring^11,12,13, previous research has employed multiple numerical values as anchors. Previous research on sequential anchoring has shown that when multiple anchors are presented sequentially, the later anchor tends to exert a stronger influence³⁹. In contrast, the Range Nudge involves a single, simultaneous presentation of multiple numerical anchors—specifically, both upper and lower limits. Thus, while both sequential anchoring and the Range Nudge utilise multiple anchors, they differ in the timing of presentation and in whether the anchors are framed as a range.
The results of this study indicate that upper/lower limits alone can be interpreted as recommended values; however, these limits are not viewed as recommended when presented within a defined range. This finding aligns with previous research indicating that lower and upper limits can be perceived as recommendations⁴⁰. The Range Nudge also altered speed evaluations. Specifically, when only the upper limit was presented, participants in the ‘60 km/h’ group tended to judge this as a low speed. These results align with prior research indicating that presenting only an upper limit shifts attention towards values above the limit and leads individuals to judge the limit as ‘low’^14,15,16. In contrast, the ‘0–60 km/h’ group effectively altered this tendency, making this speed appear higher.
Several nudges have been developed to improve safe driving^41,42 and increase handwashing behaviour^43,44. However, the Range Nudge shows that even when the semantic content of a single-limit directive and a range directive is identical—or when the explicitly stated range is narrower—framing the instruction as a range has a greater influence on behavioural adherence than presenting only an upper or lower limit. These results illustrate that the choice architecture of speed limit signs and recommended handwashing durations can significantly influence judgement and behaviour. Adopting the Range Nudge instead of limits offers two notable advantages. First, this method, which merely reframes upper/lower limits as defined ranges, can be applied to various objectives beyond speed limits and handwashing duration recommendation. For example, a doctor advising patients to reduce alcohol consumption might say, ‘You can consume up to two cans of beer per day’, leading patients to aim to drink two cans daily. However, conveying, ‘You can consume from 0 to 2 cans of beer per day’ could reduce alcohol intake. The plausible range of estimates⁴⁵ and the proximity of the anchor to the estimates⁴⁶ can impact the anchoring effect. Thus, if the range is adjustable, it should be optimised to elicit the desired behaviour depending on the target. Second, in the context of driving, simply changing the sign without altering the existing rule can reduce driving speed. For instance, both the ‘60’ and ‘0–60’ signs indicate that drivers should maintain speeds below 60 km/h. Driver education courses in Japan emphasise reducing speed when conditions require it⁴⁷. However, traditional signs cannot respond to sudden environmental changes. Recently, electronic speed limit signs have emerged, suggesting the feasibility of altering signs from ‘60’ to ‘0–60’ when lower speeds are necessary. The defined-range method is a versatile and practical solution for reducing driving speed, as it simply involves adding a zero to the speed limit sign.
The results of Study 2 indicate that altering the speed limit sign from ’60 km/h’ to ‘0–60 km/h’ led to an approximately 4.63% decrease in the number of cases in which the speed limit of 60 km/h was exceeded (from a mean of 70.38 km/h to a mean of 67.12 km/h). Studies indicate that a mere 1 km/h reduction in average speed could save approximately 2100 lives in the EU annually⁴⁸. Although this experiment was conducted under controlled conditions, the results highlight the potential for the Range Nudge to contribute to safer driving speeds and reduce speeding-related accidents. Specifically, the defined range resulted in an approximately 33.48% increase in handwashing duration in Study 4—from 28.14 s to 37.56 s—suggesting that the anchoring effect can influence adherence behaviours. Washing hands can help keep individuals healthy and prevent the spread of infections². Therefore, the Range Nudge can be associated with somewhat longer handwashing durations and may possibly lower the risk of illness.
This study did not measure how long or how often participants viewed the presented anchor values. Anchoring effects do not occur when anchors are presented subliminally⁴⁹. Anchoring effects arise from number-activated knowledge^7,8,9,10. Repeated exposure to numbers may enhance the activation of relevant knowledge, potentially intensifying the anchoring effect. Thus, the effectiveness of the Range Nudge may vary based on how long and how frequently participants are exposed to the anchor values. Further research is needed to explore these factors.
The reduction in driving speed may be due to heightened attention caused by the novelty of the road sign. However, individuals typically habituate to repeated stimuli, leading to a decrease in behavioural response over time distinct from the effects of sensory or motor fatigue⁵⁰. In Study 2, among the four courses in which the ‘0–60 km/h’ signs were repeated, no statistically significant difference in driving speed was detected (see Supplementary Table 1 in Supplementary Information for details). However, given the short duration of this study, these measures may not have fully eliminated novelty effects. Further, in other studies, the numerical representation was shown only once or twice, making it impossible to rule out novelty effects entirely. Future research should investigate the effects of the Range Nudge in studies with longer durations, multiple sessions, and varied environments.
Self-report surveys are particularly susceptible to social desirability response (SDR) bias—the tendency of participants to provide responses that are socially desirable or aligned with perceived social norms⁵¹. Thus, SDR bias may have influenced the results of Study 1, 3a, and 3b as well. In this study, only the wording—either limits or range—was altered, while the judgement process and compliance-related behaviour (e.g., recommending reduced speeding and longer handwashing duration) remained consistent across groups. Therefore, SDR bias could have had an effect regardless of whether the wording was limits or range. However, as observed in the response results from Studies 1 and 3, participants perceived the recommended driving speed to be lower when the range was used instead of limits, and they estimated the recommended handwashing duration to be longer. This difference in the perceived duration of recommendation may have led to variations in the influence of SDR bias. An SDR scale should be employed to assess the extent to which participants deliberately provide such SDR responses⁵². Given this tendency, future research should account for its potential influence when evaluating the effects of Range Nudge.
These findings are based on controlled experimental settings, and further research is needed to validate the effectiveness of the Range Nudge in real-world applications. For example, modifying road signs on actual streets or placing signage in real restrooms would provide a more accurate assessment of how effectively the Range Nudge promotes adherence behaviour. This study represents an initial step towards conducting experiments in real-world environments.
This study provides two key insights. First, the results suggest that explicitly presenting numerical limits is more effective in guiding behaviour than relying on implied semantic ranges. This underscores the critical role of numerical presentation in shaping the anchoring effect. Second, findings from experimental settings suggest that the Range Nudge, by reframing limits as defined ranges, may reduce the tendency to consider limits as recommendations and subsequently decrease non-adherence behaviours. Further research is needed to evaluate its applicability in real-world contexts.
All data are publicly available via the Open Science Framework (OSF): https://osf.io/wqjg6/.
All code is publicly accessible via the same OSF link as the data: https://osf.io/wqjg6/.
Japan Automobile Federation. English version of ‘Rules of the Road’ (Japan Automobile Federation, 2023).
Centers for Disease Control and Prevention. When and how to wash your hands. https://www.cdc.gov/clean-hands/about/index.html (2024).
Tversky, A. & Kahneman, D. Judgment under uncertainty: Heuristics and biases. Science 185, 1124–1131 (1974).
Article  PubMed  Google Scholar 
Röseler, L. et al. The open anchoring quest dataset: Anchored estimates from 96 studies on anchoring effects. J. Open Psychol. Data 10, 16 (2022).
Article  Google Scholar 
Röseler, L. et al. Correction: The open anchoring quest dataset: Anchored estimates from 96 studies on anchoring effects. J. Open Psychol. Data 12, 8 (2024).
Google Scholar 
Onuki, Y., Honda, H. & Ueda, K. What stimuli are necessary for anchoring effects to occur?. Front. Psychol. 12, 602372 (2021).
Article  PubMed  PubMed Central  Google Scholar 
Mussweiler, T. & Strack, F. Comparing is believing: A selective accessibility model of judgmental anchoring. Eur. Rev. Soc. Psychol. 10, 135–167 (1999).
Article  Google Scholar 
Mussweiler, T. & Strack, F. Hypothesis-consistent testing and semantic priming in the anchoring paradigm: A selective accessibility model. J. Exp. Soc. Psychol. 35, 136–164 (1999).
Article  Google Scholar 
Mussweiler, T. & Strack, F. The semantics of anchoring. Organ. Behav. Hum. Decis. Process 86, 234–255 (2001).
Article  Google Scholar 
Strack, F. & Mussweiler, T. Explaining the enigmatic anchoring effect: Mechanisms of selective accessibility. J. Pers. Soc. Psychol. 73, 437–446 (1997).
Article  Google Scholar 
Ames, D. R. & Mason, M. F. Tandem anchoring: Informational and politeness effects of range offers in social exchange. J. Pers. Soc. Psychol. 108, 254–274 (2015).
Article  PubMed  Google Scholar 
Gana, K. et al. Judgment of riskiness: Impact of personality, naive theories and heuristic thinking among female students. Psychol. Health 25, 131–147 (2010).
Article  PubMed  Google Scholar 
Janiszewski, C. & Uy, D. Precision of the anchor influences the amount of adjustment. Psychol. Sci. 19, 121–127 (2008).
Article  PubMed  Google Scholar 
Hohle, S. M. & Teigen, K. H. More than 50% or less than 70% chance: Pragmatic implications of single-bound probability estimates. J. Behav. Decis. Mak. 31, 138–150 (2018).
Article  Google Scholar 
Teigen, K. H. More than X is a lot: Pragmatic implicatures of one-sided uncertainty intervals. Soc. Cogn. 26, 379–400 (2008).
Article  Google Scholar 
Teigen, K. H., Halberg, A.-M. & Fostervold, K. I. More than, less than, or minimum, maximum: How upper and lower bounds determine subjective interval estimates. Behavioral Decision Making 20, 179–201 (2007).
Article  Google Scholar 
Ames, D. R. & Wazlawek, A. S. Pushing in the dark: Causes and consequences of limited self-awareness for interpersonal assertiveness. Pers. Soc. Psychol. Bull. 40, 775–790 (2014).
Article  PubMed  Google Scholar 
Guttman-Kenney, B., Leary, J. & Stewart, N. Weighing anchor on credit card debt. arXiv.org https://doi.org/10.48550/arXiv.2305.11375 (2023).
Sakaguchi, H. et al. Default effects of credit card minimum payments. J. Mark. Res. 59, 775–796 (2022).
Article  Google Scholar 
Stewart, N. The cost of anchoring on credit-card minimum repayments. Psychol. Sci. 20, 39–41 (2009).
Article  PubMed  Google Scholar 
Mori, K. et al. Effects of handwashing on Feline calicivirus removal as Norovirus surrogate. Kansenshogaku Zasshi 80, 496–500 (2006).
Article  PubMed  Google Scholar 
Kampf, G. & Ennen, J. Regular use of a hand cream can attenuate skin dryness and roughness caused by frequent hand washing. BMC Dermatol 6, 1 (2006).
Article  PubMed  PubMed Central  Google Scholar 
Thaler, R. H. & Sunstein, C. R. Nudge: Improving decisions about health, wealth, and happiness (Yale University Press, 2008).
Johnson, E. J. et al. Beyond nudges: Tools of a choice architecture. Mark. Lett. 23, 487–504 (2012).
Article  Google Scholar 
National Police Agency. Trends in enforcement of violations of the Road Traffic Act. https://www.e-stat.go.jp/en/stat-search/files?tclass=000001020603&cycle=7&year=20230 (2023).
AAA Foundation for Traffic Safety. Traffic Safety Culture Index. http://aaafoundation.org/wp-content/uploads/2023/09/202311-AAAFTS-Traffic-Safety-Culture-Index-2022.pdf (2023).
Lee, M.-S., Hong, S. J. & Kim, Y.-T. Handwashing with soap and national handwashing projects in Korea: Focus on the National handwashing Survey, 2006-2014. Epidemiol. Health 37, e2015039 (2015).
Article  PubMed  PubMed Central  Google Scholar 
Gosling, S. D. & Mason, W. Internet research in psychology. Annu. Rev. Psychol. 66, 877–902 (2015).
Article  PubMed  Google Scholar 
Podsakoff, P. M. & Podsakoff, N. P. Experimental designs in management and leadership research: Strengths, limitations, and recommendations for improving publishability. Leadersh. Q. 30, 11–33 (2019).
Article  Google Scholar 
Röseler, L. & Schütz, A. Hanging the anchor off a new ship: A meta-analysis of anchoring effects. PsyArXiv Preprints https://doi.org/10.31234/osf.io/wf2tn (2022).
Article  Google Scholar 
Hedges, L. V. Distribution theory for Glass’s estimator of effect size and related estimators. J. Educ. Behav. Stat. 6, 107–128 (1981).
Article  Google Scholar 
Cohen, J. Statistical power analysis for the behavioral sciences. 2nd ed https://doi.org/10.4324/9780203771587 (Routledge, 1988).
Goulet-Pelletier, J.-C. & Cousineau, D. A review of effect sizes and their confidence intervals, Part I: The Cohen’s d family. Tutor. Quant. Methods Psychol. 14, 242–265 (2018).
Article  Google Scholar 
Faul, F., Erdfelder, E., Lang, A.-G. & Buchner, A. G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behav. Res. Methods 39, 175–191 (2007).
Article  Google Scholar 
Lobjois, R., Faure, V., Désiré, L. & Benguigui, N. Behavioral and workload measures in real and simulated driving: Do they tell us the same thing about the validity of driving simulation?. Saf. Sci. 134, 105046 (2021).
Article  Google Scholar 
Branzi, V., Domenichini, L. & La Torre, F. Drivers’ speed behaviour in real and simulated urban roads – A validation study. Transp. Res. Part F Traffic Psychol. Behav. 49, 1–17 (2017).
Article  Google Scholar 
Watson, P. Five different “Cohen’s d” statistics for within-subject designs. MRC Cognition and Brain Sciences Unit. https://imaging.mrc-cbu.cam.ac.uk/statswiki/FAQ/tdunpaired (2017).
R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing. https://www.r-project.org/ (2022).
Bahník, Š, Houdek, P., Vrbová, L. & Hájek, J. Variations on anchoring: Sequential anchoring revisited. Judgm. Decis Mak 14, 711–720 (2019).
Article  Google Scholar 
Teigen, K. H., Halberg, A.-M. & Fostervold, K. I. Single-limit interval estimates as reference points. Appl. Cogn. Psychol. 21, 383–406 (2007).
Article  Google Scholar 
Choudhary, V., Shunko, M., Netessine, S. & Koo, S. Nudging drivers to safety: Evidence from a field experiment. Manag. Sci. 68, 4196–4214 (2021).
Article  Google Scholar 
Rubaltelli, E. et al. How to nudge drivers to reduce speed: The case of the left-digit effect. Transp. Res. Part F Traffic Psychol. Behav. 78, 259–266 (2021).
Article  Google Scholar 
Prasetyo, D. B., Sofyan, L., Muchtar, P. A. & Dewi, D. F. Nudging to handwash during the pandemic–The use of visual priming and salience. Anal. Soc. Issues Public Policy. 22, 836–856 (2022).
Article  Google Scholar 
Blackwell, C., Goya-Tocchetto, D. & Sturman, Z. Nudges in the Restroom: How Hand-Washing Can Be Impacted by Environmental Cues. SSRN https://doi.org/10.2139/ssrn.3007866 (2017).
Article  Google Scholar 
Lee, C.-Y. & Morewedge, C. K. Noise increases anchoring effects. Psychol. Sci. 33, 60–75 (2022).
Article  PubMed  Google Scholar 
Spicer, J., Zhu, J.-Q., Chater, N. & Sanborn, A. N. Perceptual and cognitive judgments show both anchoring and repulsion. Psychol. Sci. 33, 1395–1407 (2022).
Article  PubMed  Google Scholar 
The National Police Agency. Traffic safety education guidelines (in Japanese). https://www.npa.go.jp/koutsuu/kikaku/kyousoku/index.htm (2018).
Adminaité-Fodor, D. & Jost, G. Reducing speeding in Europe [PIN Flash Report 36]. European Transport Safety Council. https://etsc.eu/reducing-speeding-in-europe-pin-flash-36/ (2019).
Röseler, L. et al. Evidence against subliminal anchoring: Two close, highly powered, preregistered, and failed replication attempts. J. Exp. Soc. Psychol. 92, 104066 (2021).
Article  PubMed  Google Scholar 
Rankin, C. H. et al. Habituation revisited: An updated and revised description of the behavioral characteristics of habituation. Neurobiol. Learn. Mem. 92, 135–138 (2009).
Article  PubMed  Google Scholar 
Cerri, J., Thøgersen, J. & Testa, F. Social desirability and sustainable food research: A systematic literature review. Food Qual. Prefer 71, 136–140 (2019).
Article  Google Scholar 
van de Mortel, T. F. Faking it: Social desirability response bias in self-report research. Aust. J. Adv. Nurs. 25, 40–48 (2008).
Google Scholar 
Download references
We thank Daisuke Yamauchi, Ayumi Oka, Lu Li, and Karin Sakuragi for their assistance with video-recording the handwashing sessions and measuring the duration of handwashing. This research was funded by the Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research (B) (Y.O., JP23K28293) and the Japan Society for the Promotion of Science Grant-in-Aid for JSPS Fellows (Y.O., JP20J10404 and JP22KJ1383); the SECOM Science and Technology Foundation (K.U.); and the Japan Science and Technology Agency CREST (K.U., MJCR19A1). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Yutaro Onuki
Present address: Seijo University, Tokyo, Japan
Yutaro Onuki
Present address: Hitotsubashi University, Tokyo, Japan
The University of Tokyo, Tokyo, Japan
Yutaro Onuki & Kazuhiro Ueda
Search author on:PubMed Google Scholar
Search author on:PubMed Google Scholar
Yutaro Onuki conceptualized the study, curated the data, performed the formal analysis, acquired funding, conducted the investigation, developed the methodology, administered the project, provided resources, developed the software, validated the results, created the visualizations, and wrote the original draft. Kazuhiro Ueda contributed to the conceptualization, acquired funding, contributed to the methodology, administered the project, provided resources, supervised the project, validated the results, and reviewed and edited the manuscript.
Correspondence to Yutaro Onuki.
The authors declare no competing interests.
Communications Psychology thanks David Joachim and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Patricia Lockwood and Jennifer Bellingtier. A peer review file is available.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Reprints and permissions
Onuki, Y., Ueda, K. Range nudges enhance behavioural adherence to safety and health guidelines. Commun Psychol 3, 97 (2025). https://doi.org/10.1038/s44271-025-00276-9
Download citation
Received: 24 September 2024
Accepted: 12 June 2025
Published: 14 July 2025
Version of record: 14 July 2025
DOI: https://doi.org/10.1038/s44271-025-00276-9
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative
Communications Psychology (2026)
Collection
Advertisement
Communications Psychology (Commun Psychol)
ISSN 2731-9121 (online)
© 2026 Springer Nature Limited
Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

source