Application of Dynamic Lane-Use Control: Proposed Practices (2022)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

45   Computer-Based Comprehension Testing Methodological Considerations Using an in-person testing approach is a cost-effective way to gather data from a large number of people in different geographic areas as compared to using a simulator or closed-course approach. Respondents are typically more serious when the testing is conducted in person as opposed to online and when the respondents are being compensated for their time. In-person test administration also allows for control of the display size and color rendering, ambient noise conditions, and confirmation of demographic variables including English proficiency. This method allowed researchers to use simpler input devices than computer keyboards, which researchers have found to be more comfortable for older people and those not familiar with using computers. Computer-based testing was used to investigate the comprehension of lane- use controls because it can test more treatments than a driving simulator or closed-course study. Researchers considered using internet-based testing because of the low cost and potentially large sample size. After careful consideration, researchers rejected this idea for two reasons: (a) bias in the participant sample and (b) lack of control over display technology. The procedure began by showing the participant a scenario like, “Mary is driving in the left lane.” The subject presses a button to indicate they understood the prompt and were ready for the image to be shown. The image was then shown for five seconds, a presentation time determined through pilot testing. The subject was then given a multiple-choice question about whether or not Mary should change lanes. Figure 32 illustrates this procedure. In addition to asking multiple-choice questions for accuracy, each question also had a confidence rating attached. This combination of accuracy and confidence allowed for a more fine-grained scale of test responses. For instance, two different signs may both prompt correct answers, but if one produces more ratings of high confidence, that sign may cause less confusion than the others. Asking the questions in third person perspective is also important to avoid any misunder- standing due to language use. All drivers consider themselves to be above average, and so ques- tions that use “I” and “you” often overestimate safe and compliant behavior. A mix of names traditionally associated with males and females was used to avoid bias due to the presumed gender of the driver. Treatment Selection Researchers identified test treatments for each of the research questions selected (see Table 3 and Chapter 5). The number of treatments selected for each research question was a balance between testing several symbols of interest and having a reasonable testing length. Researchers identified the base condition, which was typically the symbol from the MUTCD. Treatment variants were then identified with the intent to compare each variant to the base condition. C H A P T E R 6

46 Application of Dynamic Lane-Use Control: Proposed Practices Depending upon the nature of the research question, the view of the signs was from the per- spective of different lanes. The view in Figure 33 was used for Research Questions 1, 4, and 5. The image for Research Question 2 (see Figure 34) included a white roadside sign that provided explanations for the symbols. Research Question 3 tested the presence of a CMS (see the exam- ple in Figure 35). Figure 36 shows the image for Research Questions 9 and 11 where the driver is in Lane 2. The driver was in Lane 3 for Research Questions 10, 12, and 13 (see Figure 37). Figure 38 shows the image used to test a shoulder-mounted sign. The background images did Button Press Presentation for 5 sec Button Press Answer In the next photo, Mary is in Lane 4. Press 7 to continue. Should Mary stay in her lane (Lane 4)? 1) Yes, that lane is open. 2) Yes, but she should be prepared for the lane to be closed further down the road. 3) No, that lane is closed right there. 4) No, that symbol will change to red in a few seconds. Figure 32. General computer-based testing procedure. Figure 33. Example of the image for Research Questions 1, 4, and 5. Figure 34. Example of the image for Research Question 2.

Computer-Based Comprehension Testing 47   Figure 35. Example of the image for Research Question 3 for Treatment 10 with supplemental CMS showing STALLED VEHICLE AHEAD. Figure 36. Example of the image for Research Questions 9 and 11. Figure 37. Example of the image for Research Questions 10, 12, and 13.

48 Application of Dynamic Lane-Use Control: Proposed Practices not have any traffic shown. This was because the appearance of other traffic using the lane may influence the person’s understanding of whether or not the lane is open. Experimental Design For each version, the test questions were presented in a partially randomized order that minimized opportunities for learning effects to influence responses. This is because some of the test questions include information that will reveal the meanings of some of the signs. Thus, those questions appeared near the end of the presentation order. The order of presentation was quasi-randomized with the constraint that a control and its variant could not appear in succession in the order of testing. Additional details about the test question order are available in Appendix D. Study Approach Data were collected in four cities (Dallas, Texas; Columbus, Ohio; Las Vegas, Nevada; and Raleigh, North Carolina) representing a range of sizes, regions of the country, and use of color CMSs. All recruiting and experimental procedures were reviewed and approved by the Texas A&M University Office of Human Subjects Protection and the Institutional Review Board and were also reviewed by the National Academies Institutional Review Board. In total, 320 licensed drivers participated in the computer-based testing. The number was approximately evenly split across the four testing locations. A summary of the demographic characteristics of the participants is provided in Appendix D. Recruiting efforts were directed toward achieving a mix of ages and genders similar to that of licensed drivers in each state. Researchers recruited participants by contacting individuals who had participated in past research studies in each location, and through social media posts, posted flyers, and word of mouth. Participants were paid $50 for completing the test. The testing session took approxi- mately 1 hour, including time for instruction, practice, testing, and debriefing. Participants were scheduled at 10-minute intervals to allow experimenters time to pro- vide individual instructions to each participant in a room separate from where the testing workstations were located. Participants were greeted, completed the consent form, and pro- vided instructions. A workstation was in the instruction room to provide participants an Figure 38. Example of the image for Research Question 20, shoulder-mounted sign proposed for use in England as ATM Lite application.

Computer-Based Comprehension Testing 49   opportunity to complete a short practice test to become familiar with the button boxes and question format. Statistical Evaluation Approach Fisher’s exact test was used, which is a statistical test used for categorical variables such as the ones used in computer-based testing. Fisher’s exact test was conducted to explore if the distributions between responses for a pair of treatments differed. Fisher’s exact was selected over chi-squared because there were cell sizes with less than five responses for several of the treatments. A p-value less than 0.05 was considered to be statistically significant. Research Questions, Treatments, and Results The following sections present an overview of the situation being investigated for each research question. The results from the computer-based test are presented within a figure and are then followed by a discussion as to how those results could potentially impact practices. The order of the research questions is not numerical; rather, they are presented so that similar questions follow each other. Appendix D contains additional details on the treatments and the paired comparisons. Research Question 1: Does including words with red, yellow, and/or green symbols improve comprehension? Overview Supplemental text (e.g., CAUTION, 1 MILE, MERGE, and CLOSED) has been displayed in conjunction with some symbols. This question focused on the comprehension changes when the symbol sign is accompanied by text. An example of the freeway image seen by the participants for this question is shown in Figure 33. Figure 39 shows the treatment tested along with provid- ing the test questions. Results Figure 39 provides the percentage of responses for each treatment tested within Research Question 1. It also provides the distribution of responses to the confidence question along with the number of participants who saw that treatment. Table 4 lists the paired comparison tests. It also shows the findings of the Fisher’s exact tests conducted to determine whether the distribu- tions of responses for the treatments differed among the unpaired data. The shaded cells indicate comparisons that were significantly different at the 0.05 level in the statistical test. Discussion Some agencies have adopted the practice of adding text below or above each symbol to indi- cate its meaning. Adding this text necessarily makes the symbol smaller or requires a larger display board. Adding this text increases the information units in the display as well, which could increase driver reading time, particularly for those whose first language is not English. The results of the study showed the benefit of adding text for some symbols and not others. The results showed no benefit of adding words to the symbols for open and closed (i.e., green arrow and red X). Adding words to the yellow symbols, however, did affect the driver’s under- standing of the symbol. Adding the word MERGE prompted more drivers to interpret the downward yellow arrow as indicating an immediate lane closure when compared to a diagonal arrow with no text.

50 Application of Dynamic Lane-Use Control: Proposed Practices Ta Lane1 Lane 2 Lane 3 Lane 4 Responses (%) to Test Questionb Responses (%) to Confidence Questionc Nd 1 2 3 4 1 2 3 4 5 1 1 12 88 NAe 1 2 7 17 73 128 2 0 11 89 NA 0 2 3 9 86 64 3 2 82 9 8 0 6 20 25 48 320 4 0 71 19 10 0 5 16 25 54 223 5 0 91 4 4 1 3 9 21 67 320 6 7 60 25 8 1 11 22 29 37 320 7 2 85 7 6 1 4 13 26 57 225 8 95 3 2 0 0 0 3 8 89 63 9 97 3 0 0 0 0 0 6 94 63 aT = Treatment bTest Question: Should <name> stay in the right lane (Lane 4)? 1) Yes, that lane is open. 2) Yes, but she should be prepared for the lane to be closed farther down the road. 3) No, that lane is closed right there. 4) No, that symbol will change to red in a few seconds. cConfidence Question: How confident are you that your answer is correct? 1) Not at all confident, basically guessing. 2) 50% confident, I could narrow it down to two of the options. 3) 75% confident, pretty sure I’m right. 4) 95% confident, slight chance I’m wrong. 5) 100% confident. dN = Sample Size eNA = Not Applicable Figure 39. Responses for Research Question 1. Research Question Comparison Fisher’s Exact p-value Statistically Significant Difference in Distribution Research Question 1 T1 vs. T2 1.000 No T3 vs. T4 0.001 Yes T3 vs. T5 0.005 Yes T6 vs. T7 0.000 Yes T8 vs. T9 1.000 No Table 4. Fisher’s exact results for Research Question 1. e 2019 FHWA memorandum stated a steady yellow X signal indication means “that the lane it is over is about to be closed to trac, and road users should prepare to vacate the lane.” In other words, the yellow X is used like a clearance interval at a signalized intersection to warn of an impending red indication e results of this study show that only 6%–8% of respondents interpret the symbol that way, indicated by selecting Option 4 (No, that symbol will change to red in a few seconds) to the test question of whether the example driver should stay in the lane or not. e results showed that 25% of test subjects thought a yellow X meant the lane was closed at that location. Only a slight majority (60%) thought the yellow X meant that drivers should prepare for the lane to be closed farther down the road. e downward diagonal yellow arrow, on the other hand, produced 82% of respondents understanding that the lane closure was down- stream. e research results recommend using a downward diagonal yellow arrow to indicate impending lane closure rather than a yellow X.

Computer-Based Comprehension Testing 51   Adding the text 1 MILE to both the yellow X and the downward diagonal yellow arrow that were included as variations did better communicate that the lane was not closing at the location of the sign. As will be discussed in Research Question 10, adding the word SLOW is useful to convey that the lane is open, but trac is moving slowly in it. erefore, if an agency wishes to use the MUTCD yellow X symbol to convey impending lane closure, the research recommends adding a distance legend to the display. Research Question 5: What is the best symbol to communicate that the lane is closed ahead, and the driver needs to vacate the lane? Overview Figure 40 shows the treatments considered within Research Question 5 along with the test questions. For this research question, the symbols were displayed on a background image simi- lar to that shown in Figure 33 (i.e., with an overhead sign bridge). e symbols were displayed in a steady state (i.e., no ashing or scrolling). Treatment 6 is the MUTCD symbol intended for reversible lanes to indicate an impending change to red X but is used in some ATM applications to indicate both “lane open – but slow down” and “lane closed ahead – merge” based on recommendations in the MUTCD. Treat- ment 1 is the MUTCD baseline treatment for Lane 4 being closed (i.e., now, not farther ahead). Ta Lane1 Lane 2 Lane 3 Lane 4 Responses (%) to Test Questionb Responses (%) to Confidence Questionc Nd 1 2 3 4 1 2 3 4 5 1 1 12 88 NAe 1 2 7 17 73 128 3 2 82 9 8 0 6 20 25 48 320 4 0 71 19 10 0 5 16 25 54 223 5 0 91 4 4 1 3 9 21 67 320 6 7 60 25 8 1 11 22 29 37 320 7 2 85 7 6 1 4 13 26 57 225 13 2 85 6 7 3 2 28 27 40 129 14 0 75 17 8 0 9 19 28 44 64 15 3 70 19 8 0 14 30 19 37 63 aT = Treatment bTest Question: Should <name> stay in the right lane (Lane 4)? 1) Yes, that lane is open. 2) Yes, but she should be prepared for the lane to be closed farther down the road. 3) No, that lane is closed right there. 4) No, that symbol will change to red in a few seconds. cConfidence Question: How confident are you that your answer is correct? 1) Not at all confident, basically guessing. 2) 50% confident, I could narrow it down to two of the options. 3) 75% confident, pretty sure I’m right. 4) 95% confident, slight chance I’m wrong. 5) 100% confident. dN = Sample Size eNA = Not Applicable Figure 40. Responses for Research Question 5.

52 Application of Dynamic Lane-Use Control: Proposed Practices The remaining treatments shown in Figure 40 are experimental treatments for indicating that a driver should merge out of Lane 4. This research question was used to determine which of the symbols shown is best understood to indicate that the lane is closed ahead. To investigate the understanding of a symbol with text added, two treatments with such combinations are included. Treatment 3 (plain downward diagonal yellow arrow) will be compared to versions with added text MERGE (Treatment 4) and 1 MILE (Treatment 5). The yellow X with no text (Treatment 6) is compared to a version with text 1 MILE (Treatment 7). Results Figure 40 provides the percentage of responses for each treatment tested within Research Question 5. It also provides the distribution of responses to the confidence question along with the number of participants who saw that treatment. Table 5 shows the paired comparisons along with the findings of the Fisher’s exact tests conducted to determine whether the distributions of responses for the treatments differed among the unpaired data. Discussion Research Question 5 built upon symbols tested in Research Question 1 to include more arrow designs that were found in use by the state-of-the-practice review. The upward diagonal yellow arrow, chevron symbols, and horizontal arrow (Treatments 13–15) were compared to the yellow X and downward diagonal yellow arrow with and without text presented in Research Question 1. In addition to the discussion presented in Research Question 1, the treatment variations added here demonstrate that a diagonal arrow pointing either upward or downward does a better job than a yellow X of communicating that the lane is closing farther down the road and not at the sign location. The steady, three-chevron symbol also performed similarly to the upward and downward diagonal arrows. Research Question 10: What is the best symbol to communicate caution or that slow traffic is ahead when the lane is still open? Overview For this question, the symbols were displayed on the sign bridge as shown in Figure 37. The questions asked, and the treatments evaluated are shown in Figure 41. Treatment 21 is the MUTCD baseline treatment for Lane 3 being closed ahead. At this time, there is no MUTCD symbol that indicates that the lane is open ahead, but the driver should proceed with caution or slow down. Some agencies are using a down yellow arrow (Treatment 22) to communicate that a lane is open but moving slowly. Therefore, all the participants saw a Research Question Comparison Fisher’s Exact p-value Statistically Significant Difference in Distribution Research Question 5 T6 vs. T1 0.000 Yes T6 vs. T3 0.000 Yes T6 vs. T13 0.000 Yes T6 vs. T14 0.034 Yes T6 vs. T15 0.477 No T4 vs. T5 0.000 Yes T5 vs. T7 0.123 No T3 vs. T13 0.818 No T3 vs. T14 0.211 No Table 5. Fisher’s exact results for Research Question 5.

Computer-Based Comprehension Testing 53   comparison between the yellow X (Treatment 21, the MUTCD baseline) and the yellow down arrow. The other treatments are all experimental treatments intended to meet the need to com- municate caution. Results Figure 41 provides the percentage of responses for each treatment tested within Research Question 10. It also provides the distribution of responses to the confidence question along with the number of participants who saw that treatment. Table 6 shows the pairs of treatments compared along with the findings of the Fisher’s exact tests conducted to determine whether the distributions of responses for the treatments differed among the unpaired data. Ta Lane 1 Lane 2 Lane 3 Lane 4 Responses (%) to Test Questionb Responses (%) to Confidence Questionc N d 1 2 3 4 5 1 2 3 4 5 21 2 25 57 9 7 2 11 21 33 34 320 22 6 54 37 2 2 1 12 21 30 37 320 23 6 64 23 6 0 1 6 17 30 46 81 24 13 81 1 4 1 0 8 9 26 58 80 25 9 78 13 1 0 0 4 10 33 54 80 26 5 41 48 4 3 1 9 22 32 35 79 aT = Treatment bTest Question: Should <name> stay in her lane (Lane 3)? 1) Yes, that lane is open. 2) Yes, but she should slow down. 3) Yes, but she should be prepared for the lane to be closed farther down the road. 4) No, that lane is closed right there. 5) No, that symbol will change to red in a few seconds. cConfidence Question: How confident are you that your answer is correct? 1) Not at all confident, basically guessing. 2) 50% confident, I could narrow it down to two of the options. 3) 75% confident, pretty sure I’m right. 4) 95% confident, slight chance I’m wrong. 5) 100% confident. dN = Sample Size Figure 41. Responses for Research Question 10. Research Question Comparison Fisher’s Exact p-value Statistically Significant Difference in Distribution Research Question 10 T21 vs. T22 0.000 Yes T21 vs. T23 0.000 Yes T21 vs. T24 0.000 Yes T21 vs. T25 0.000 Yes T21 vs. T26 0.015 Yes T22 vs. T23 0.025 Yes T22 vs. T24 0.000 Yes T22 vs. T25 0.000 Yes T22 vs. T26 0.124 No T23 vs. T25 0.089 No T24 vs. T25 0.020 Yes Table 6. Fisher’s exact results for Research Question 10.

54 Application of Dynamic Lane-Use Control: Proposed Practices Discussion This research question gets to the heart of the ATM operation of trying to warn drivers of a queue ahead; the driver can stay in the lane, but traffic ahead is slowed. The MUTCD currently does not address this operational state. The test responses to this question show that, for the participants in this study, the downward yellow arrow communicates a different message (slow down) than the yellow X (be prepared to leave lane). Even when the word SLOW is added to the yellow X, drivers still believe the X means “be prepared to leave lane.” The downward yellow arrow communicated that the lane was open both with and without the SLOW legend added. The downward green arrow with the SLOW legend performed well, but 13% of the people incor- rectly chose Option 1 (The lane is open)—a dangerous misinterpretation. Another 13% of the people incorrectly chose Option 3 (Prepare for the lane to be close farther down the road), which is a safer misinterpretation. For this reason, the research recommends the addition of the downward yellow arrow to the MUTCD set of lane-use control symbols to be used for an end-of-queue warning. Further, researchers recommend the use of the legend SLOW in conjunction with the downward yellow arrow. If an agency wishes to use the current MUTCD lane-use control symbols, this research sug- gests that the legend SLOW should be added to the yellow X to increase the number of people understanding that the lane is not going to be closed downstream. The word SLOW performed significantly better than the word CAUTION in conveying this message for the down arrow and presumably would also be better when used in conjunction with the yellow X (CAUTION was only tested with the downward yellow arrow). Research Question 9: For center lane closures, what is the best symbol for indicating the need to merge either left or right? Overview For this question, the symbols were displayed on a background image similar to that shown in Figure 36. The treatments tested are shown in Figure 42. Treatment 16 is the MUTCD baseline treatment for Lane 2 being closed ahead. It does not provide any directional information indi- cating which lane the driver should merge into. Treatments 17 to 20 are experimental treatments for indicating that a driver should merge out of Lane 2. Treatment 17 appears to indicate that the driver should merge into Lane 3, while the other experimental treatments may be interpreted to mean that the driver may move into either Lane 1 or Lane 3. This question was used to deter- mine which of the symbols shown is best understood to indicate that the driver has the option to choose either neighboring lane to merge into when one of the center lanes is closed ahead. Results Figure 42 provides the percentage of responses for each treatment tested within Research Question 9. It also provides the distribution of responses to the confidence question along with the number of participants who saw that treatment. Table 7 shows the pairs being compared along with the findings of the Fisher’s exact tests conducted to determine whether the distribu- tions of responses for the treatments differed among the unpaired data. Discussion The treatment variations tested in this research question showed that a yellow X is com- municating the need to merge out of the lane. The alternates that would be used for center lane closures to indicate merging into either adjacent lane all performed identically to each other and the yellow X. The steady downward diagonal arrow pointing in a single direction

Computer-Based Comprehension Testing 55   caused the majority (53%) of people to think they could merge only into the lane adjacent to the right. Research Question 4: Does scrolling or flashing some or all of the message elements affect comprehension? Overview For this question, the symbols were displayed on a background image similar to that shown in Figure 33 (i.e., with an overhead sign bridge). Figure 43 provides the treatments tested along with the test questions. Note that some participants were shown Treatment 3 as part of Research Question 1. Those participants were not shown Treatment 3 a second time. Results Figure 43 provides the percentage of responses for each treatment tested within Research Question 4. It also provides the distribution of responses to the confidence question along with the number of participants who saw that treatment. Table 8 shows the paired comparisons along with the findings of the Fisher’s exact tests conducted to determine whether the distributions of responses for the treatments differed among the unpaired data. Figure 42. Responses for Research Question 9. Ta Lane 1 Lane 2 Lane 3 Lane 4 Responses (%) to Test Questionb Responses (%) to Confidence Questionc N d 1 2 3 1 2 3 4 5 16 3 7 90 0 4 9 22 65 320 17 5 53 42 2 6 6 28 57 81 18 4 3 94 0 5 4 9 83 80 19 1 1 98 3 1 3 10 84 80 20 4 1 95 0 0 4 10 86 79 aT = Treatment bTest Question: <name> is in Lane 2 and believes that Lane 2 is closed ahead. What should he do? 1) Merge into the lane to the left (Lane 1). 2) Merge into the lane to the right (Lane 3). 3) Merge into either Lane 1 or Lane 3. cConfidence Question: How confident are you that your answer is correct? 1) Not at all confident, basically guessing. 2) 50% confident, I could narrow it down to two of the options. 3) 75% confident, pretty sure I’m right. 4) 95% confident, slight chance I’m wrong. 5) 100% confident. dN = Sample Size Table 7. Fisher’s exact results for Research Question 9. Research Question Comparison Fisher’s Exact p-value Statistically Significant Difference in Distribution Research Question 9 T16 vs. T17 0.000 Yes T16 vs. T18 0.322 No T16 vs. T19 0.088 No T16 vs. T20 0.116 No

56 Application of Dynamic Lane-Use Control: Proposed Practices Discussion The results of the test indicate that flashing and scrolling seem to communicate more urgency compared to steady indications. More people understood the scrolling chevron (Treatment 11) and flashing downward diagonal arrow (Treatment 12) to indicate that the lane was closed at the sign location when compared to a steady diagonal arrow. With the increased prevalence of flashing yellow arrows at signalized intersections for permitted left-turn phases, it is not sur- prising that 14% of subjects thought a flashing downward diagonal arrow meant the signal was about to change to red. While some agencies are using flashing and scrolling to increase conspicuity, it appears from the testing that this motion is changing the meaning of the symbol for drivers. Research Question 11: When an adjacent lane is closed, what is the best symbol for indicating the potential presence of emergency vehicles in the blocked lane? Overview For this research question, the symbols were displayed on an overhead sign bridge with the driver in Lane 2 (see background image example in Figure 36). The questions asked and the Ta Lane 1 Lane 2 Lane 3 Lane 4 Responses (%) to Test Questionb Responses (%) to Confidence Questionc N d 1 2 3 4 1 2 3 4 5 11 0 75 18 7 1 7 12 28 52 145 3 2 82 9 8 0 6 20 25 48 320 12 1 67 18 14 1 10 24 23 42 175 aT = Treatment bTest Question: Should <name> stay in the right lane (Lane 4)? 1) Yes, that lane is open. 2) Yes, but she should be prepared for the lane to be closed farther down the road. 3) No, that lane is closed right there. 4) No, that symbol will change to red in a few seconds. cConfidence Question: How confident are you that your answer is correct? 1) Not at all confident, basically guessing. 2) 50% confident, I could narrow it down to two of the options. 3) 75% confident, pretty sure I’m right. 4) 95% confident, slight chance I’m wrong. 5) 100% confident. dN = Sample Size Figure 43. Responses for Research Question 4. Research Question Comparison Fisher’s Exact p-value Statistically Significant Difference in Distribution Research Question 4 T3 vs. T11 0.020 Yes T3 vs. T12 0.001 Yes Table 8. Fisher’s exact results for Research Question 4.

Computer-Based Comprehension Testing 57   treatments evaluated are shown in Figure 44. The image had Lane 1 with a red X, indicating that the lane is closed. The treatments are similar to Research Question 10 with the difference that the adjacent lane is now shown as being closed. Also similar to Research Question 10, all the participants saw the primary comparison of yellow X to the downward yellow arrow, except in the Lane 2 position rather than the Lane 3 position, being used in Research Question 10. Treatment 27 is the MUTCD baseline treatment for Lane 1 being closed and Lane 2 being closed ahead. At this time, there is no MUTCD symbol that could be used to indicate that Lane 2 is open ahead, but the driver should proceed with caution or slow down. The other treatments shown in Figure 44 are experimental treatments intended to meet that need. It is not known if drivers would use caution adjacent to a closed lane without some other indication to do so. The flashing downward yellow arrow (Treatment 31) was added because it was observed in Minnesota in August 2018. All participants saw Treatment 27 and Treatment 28 to investigate if the downward yellow arrow symbol is similar in meaning to the downward green arrow or if the yellow X also conveys caution. The participants also saw Treatment 29, 30, 31, 32, or 33 compared to both the yellow X and the downward yellow arrow. Results Figure 44 provides the percentage of responses for each treatment tested within Research Question 11. It also provides the distribution of responses to the confidence question along with the number of participants who saw that treatment. Table 9 shows the paired comparisons and the findings of the Fisher’s exact tests conducted to determine whether the distributions of responses for the treatments differed among the unpaired data. Ta Lane 1 Lane 2 Lane 3 Lane 4 Responses (%) to Test Questionb Responses (%) to Confidence Questionc Nd 1 2 3 1 2 3 4 5 27 3 45 53 3 13 25 26 33 320 28 7 37 56 2 10 18 36 34 320 29 3 42 55 2 13 23 27 36 64 30 6 36 58 2 13 22 33 31 64 31 6 35 58 8 14 31 26 22 65 32 44 42 14 2 2 8 17 72 64 33 19 35 46 2 8 24 29 38 63 aT = Treatment bTest Question: What is the sign over Lane 2 telling you to do? 1) Stay in the lane. 2) Move out of the lane because there may be emergency vehicles in Lane 1. 3) Stay in the lane but be prepared to move if the sign in Lane 2 turns to a red X. cConfidence Question: How confident are you that your answer is correct? 1) Not at all confident, basically guessing. 2) 50% confident, I could narrow it down to two of the options. 3) 75% confident, pretty sure I’m right. 4) 95% confident, slight chance I’m wrong. 5) 100% confident. dN = Sample Size Figure 44. Responses for Research Question 11.

58 Application of Dynamic Lane-Use Control: Proposed Practices Discussion Research Question 11 examined treatments that should be used over lanes adjacent to closed lanes. The test results showed that neither the addition of words nor the flashing arrow symbol affected drivers’ choices. A downward green arrow did prompt more people to select “stay in lane.” These treatments intended to get people to slow down in a lane adjacent to a closed lane, and none of the treatments did a good job of achieving these results. The answer choices for this set of questions were different from others. This was done to add some variety. In retro- spect, a choice that clearly stated “stay in lane but slow down” would have been appropriate for this question. Research Question 12: Should flashing or other conspicuity- enhancing methods be used for a caution messages? Overview For this question, the symbols were displayed on a sign bridge with the driver in Lane 3 (see Figure 37 for an example view). The questions asked and the treatments evaluated are shown in Figure 45. For this question, Lane 4 had a red X, indicating that the lane is closed. This question is very similar to Research Question 11 in that it addresses driver response in a lane adjacent to a closed lane. Again, Treatment 34 is the MUTCD baseline treatment for Lane 4 being closed and Lane 3 being closed ahead. Treatments 35, 36, and 37 are intended to indicate that the driver should use caution in Lane 3. Results Figure 45 provides the percentage of responses for each treatment tested within Research Question 12. It also provides the distribution of responses to the confidence question along with the number of participants who saw that treatment. Table 10 shows the pairs compared and the findings of the Fisher’s exact tests. Discussion The state-of-the-practice review showed that some agencies were flashing symbols to increase their conspicuity. The results of this study did show that flashing seemed to increase a sense of urgency in exiting the lane, with more participants opting to get out of the lane immediately in response to the flashing. Adding the SLOW legend here did increase the number of people indicating they would stay in the lane and slow down. This finding concurs with that in Research Question 10 where the closed lane was the far right lane. Note that this set of questions had slightly different response options than other sets. This is another piece of evidence—asked Research Question Comparison Fisher’s Exact p-value Statistically Significant Difference in Distribution Research Question 11 T27 vs. T28 0.007 Yes T27 vs. T29 0.799 No T27 vs. T30 0.138 No T27 vs. T31 0.126 No T27 vs. T32 0.000 Yes T27 vs. T33 0.000 Yes T28 vs. T29 0.470 No T28 vs. T30 0.976 No T28 vs. T31 0.953 No T28 vs. T32 0.000 Yes T28 vs. T33 0.018 Yes Table 9. Fisher’s exact results for Research Question 11.

Computer-Based Comprehension Testing 59   differently and applied to a lane adjacent to a closed lane—that supports the recommendation that adding SLOW to a symbol encourages people to stay in the lane but slow down. Adding flashing icons to the corner of the symbol display did not have any effect, which may be an artifact of the image size. The participants may have responded based on seeing the large X and pressed the button before the small flashing icons were noticed. Research Question 13: How do drivers interpret the various resting modes (e.g., blank or dark, green arrow)? Overview Some agencies turn off the lane-control signals when traffic is moving smoothly, and no special- use lanes are in use. Other agencies display a downward green arrow over each lane during these normal operations. Researchers believed that a blank sign may cause some confusion to drivers who may think the signs have malfunctioned or lost power if they are blank. This research ques- tion addresses this topic. The base freeway image used in the testing is shown in Figure 37. In this application, the sign’s resting state means that the lanes are open and that there is no lane-use information to report to motorists. The questions asked and the treatments evaluated are shown in Figure 46. For Research Question 13, Treatment 8 will be compared to Treatment 38. Ta Lane 1 Lane 2 Lane 3 Lane 4 Responses (%) to Test Questionb Responses (%) to Confidence Questionc Nd 1 2 3 4 1 2 3 4 5 34 11 14 21 54 1 10 20 29 40 320 35 1 5 19 75 1 6 19 31 44 144 36 3 14 21 63 3 8 21 21 48 80 37 6 11 42 41 0 9 21 32 38 96 aT = Treatment bTest Question: <name> is in Lane 3 and believes that Lane 4 is closed. What should he do? 1) Stay in Lane 3 and continue driving normally. 2) Stay in Lane 3 and continue driving normally while scanning for hazards in the lanes ahead. 3) Stay in Lane 3 and slow down while scanning for hazards in the lanes ahead. 4) Merge out of Lane 3 into Lane 2. cConfidence Question: How confident are you that your answer is correct? 1) Not at all confident, basically guessing. 2) 50% confident, I could narrow it down to two of the options. 3) 75% confident, pretty sure I’m right. 4) 95% confident, slight chance I’m wrong. 5) 100% confident. dN = Sample Size Figure 45. Responses for Research Question 12. Research Question Comparison Fisher’s Exact p-value Statistically Significant Difference in Distribution Research Question 12 T34 vs. T35 0.000 Yes T34 vs. T36 0.080 No T34 vs. T37 0.001 Yes Table 10. Fisher’s exact results for Research Question 12.

60 Application of Dynamic Lane-Use Control: Proposed Practices Results Figure 46 provides the percentage of responses for each treatment tested within Research Ques- tion 13. It also provides the distribution of responses to the confidence question along with the number of participants who saw that treatment. The Fisher’s exact test found a significant differ- ence (p-value = 0.000) in the distributions of responses between Treatment 8 and Treatment 38. Discussion The state-of-the-practice review showed that several agencies blank their signs during normal operations. This could be interpreted by drivers to mean that the signs are not receiving traffic data or have lost power. The results of this question clearly show that a display of downward green arrows (Treatment 8) more clearly conveys that traffic is moving normally than all dark signs (Treatment 38). Research Question 20: Would the comprehension of a lane-use symbol be similar for a shoulder-mounted sign as for an overhead sign bridge? Overview Overhead sign structures are expensive to install and may require lane closures for mainte- nance. Some agencies are experimenting with the use of shoulder-mounted CMSs illustrating a symbol per lane in place of overhead structures (called ATM Lite in some areas). For this question, the comparison was between a participant’s response to similar symbols being used on a left-side shoulder-mounted sign as an overhead sign bridge. The image seen by the partici- pant for the left-side shoulder-mounted sign is shown in Figure 38. The treatments evaluated are similar to some of those in Research Question 5, which were on an overhead sign bridge. Figure 47 lists the symbols included within Research Question 20. The symbols were displayed in a steady state (i.e., no flashing or scrolling). Results Figure 47 provides the percentage of responses for each treatment tested within Research Question 20. Note that the responses were grouped into two general categories of Yes (the lane Ta Lane 1 Lane 2 Lane 3 Lane 4 Responses (%) to Test Question Responses (%) to Confidence Questionc N e Q 1 2 3 1 2 3 4 5 8 Q1 b 63 3 34 3 4 10 22 61 320 38 Q2 d 25 20 55 4 8 11 23 53 320 aT = Treatment bTest Question 1 (Q1): <Name> is driving in Lane 3. What is happening? 1) Traffic is moving normally; the signs are in default mode. 2) The signs have not been updated with current traffic information. 3) Not sure, either of the choices above could be true. cConfidence Question: How confident are you that your answer is correct? 1) Not at all confident, basically guessing. 2) 50% confident, I could narrow it down to two of the options. 3) 75% confident, pretty sure I’m right. 4) 95% confident, slight chance I’m wrong. 5) 100% confident. dTest Question 2 (Q2): <Name> is driving in Lane 3. What is happening? 1) Traffic is moving normally, so no message is needed. 2) The signs are broken/malfunctioning. 3) Not sure, either of the choices above could be true. eN = Sample Size Figure 46. Responses for Research Question 13.

Computer-Based Comprehension Testing 61   Ta Locb Lane1 Lane 2 Lane 3 Lane 4 Responses (%) to Test Question Responses (%) to Confidence Question c N d Q No Yes 1 2 3 4 5 1 Over Q1e 88 13 1 2 7 17 73 128 40 Road-side Q2f 84 16 2 6 3 16 73 64 3 Over Q1e 17 83 0 6 20 25 48 320 41 Road-side Q2f 9 91 5 6 16 30 44 64 6 Over Q1e 33 67 1 11 22 29 37 320 39 Road-side Q2f 17 83 6 5 23 27 39 64 13 Over Q1e 13 87 3 2 28 27 40 129 42 Road-side Q2f 15 85 3 11 26 25 35 65 15 Over Q1e 27 73 0 14 30 19 37 63 43 Road-side Q2f 14 86 0 10 24 32 35 63 aT = Treatments bLoc = Location of signs: Over = over the lanes and Roadside = on the roadside cConfidence Question: How confident are you that your answer is correct? 1) Not at all confident, basically guessing. 2) 50% confident, I could narrow it down to two of the options. 3) 75% confident, pretty sure I’m right. 4) 95% confident, slight chance I’m wrong. 5) 100% confident. dN = Sample Size eTest Question 1 (Q1): Should <name> stay in the right lane (Lane 4)? Answers are bundled as follows: Yes = Yes, that lane is open and Yes, but she should be prepared for the lane to be closed farther down the road. No = No, that lane is closed right there and No, that symbol will change to red in a few seconds (when relevant). fTest Question 2 (Q2): Should <name> stay in the right lane (Lane 4)? Answers are bundled as follows: Yes = Yes, that lane is open; Yes, but she should be prepared for the lane to be closed farther down the road; and Yes, but she should slow down. No = No, that lane is closed right there and No, that symbol will change to red in a few seconds (when relevant). Figure 47. Responses for Research Question 20. Note that T1 to T15 are overhead signs, while T39 to T43 are shoulder-mounted signs. (See example in Figure 38). is open) and No (lane is closed or closing) because the response presented for the overhead signs was slightly dierent as compared to the roadside signs. Figure 47 also provides the distribution of responses to the condence question along with the number of participants who saw that treatment. Table 11 shows paired comparisons and the ndings of the Fisher’s exact tests con- ducted to determine whether the distributions of responses for the treatments diered among the unpaired data. Research Question Comparison Fisher’s Exact p-value Statistically Significant Difference in Distribution Research Question 20 T39 vs. T6 0.011 Yes T40 vs. T1 0.655 No T41 vs. T3 0.185 No T42 vs. T13 0.666 No T43 vs. T15 0.122 No Table 11. Fisher’s exact results for Research Question 20.

62 Application of Dynamic Lane-Use Control: Proposed Practices Discussion The distribution of “Yes, stay in lane” as compared to “No, leave lane” were similar whether the symbols were overhead or were roadside with one exception. The yellow X was associated with more drivers saying “stay in lane” when the symbol was shown on the roadside as compared to overhead. Note that the sample size for the roadside questions was 64 while it was 320 for the overhead questions. Agencies wishing to reduce installation and maintenance costs may want to convey DLUC information using a shoulder-mounted sign rather than an overhead sign. The results of the study show that the roadside signs performed similarly to overhead signs within the context of the four-lane freeway tested here and for most symbols tested. More research is recommended to test the general application of roadside lane-use signs on different types of facilities. Research Question 2: Do supplemental static signs that explain the meaning of symbols improve comprehension? Overview Research Question 2 was presented as the final question at the end of the testing. This is because researchers did not want the explanatory signs to provide information that participants may use to respond to other research questions. The image for the test questions is shown in Figure 34 and was scaled appropriately for the display screens used in the testing. The shoulder sign is based on one in use by VDOT. Results Figure 48 provides the percentage of responses for each treatment tested within Research Question 2. It also provides the distribution of responses to the confidence question along with the number of participants who saw that treatment. Ta Lane 1 Lane 2 Lane 3 Lane 4 Responses (%) to Test Question Responses (%) to Agreement Questionc Ne Q 1 2 3 4 1 2 3 4 5 44 Q1b 7 9 19 65 2 2 16 34 46 320 Q2d 22 32 33 14 NAf NA NA NA NA 320 aT = Treatment bTest Question 1 (Q1): How helpful do you think the white sign is for drivers to understand the symbols on the overhead signs? 1) Not at all helpful. 2) A little helpful. 3) Somewhat helpful. 4) Very helpful. cAgreement Question: How many other drivers in your community do you think would agree with you? 1) Nearly none. 2) Less than half. 3) About half. 4) More than half. 5) Nearly all drivers. dTest Question 2 (Q2): Overhead signs like the ones here often appear every 0.5 mile or so. How frequently do you think a white sign that explains the symbols used overhead would need to appear? 1) Next to every overhead sign. 2) Next to the first overhead sign in a series. 3) Next to every other overhead sign. 4) I don’t think these white signs are needed at all. eN = Sample Size fNA = Not Applicable Figure 48. Responses for Research Question 2. Note that a static, ground-mounted sign was included in addition to the DLUC signals.

Computer-Based Comprehension Testing 63   Figure 49. Responses for Research Question 3. Ta Lane1 Lane 2 Lane 3 Lane 4 Responses (%) to Test Question Responses (%) to Confidence Questionc Ne Q 1 2 3 4 1 2 3 4 5 10 w/ CMS Q1b 5 73 11 11 2 12 19 33 34 160 Q2d 25 75 NAf NA NA NA NA NA NA 160 10 w/o CMS Q1 3 80 8 10 1 7 29 28 35 160 aT = Treatment bTest Question 1 (Q1): Should <name> stay in his lane (Lane 3)? 1) Yes, that lane is open. 2) Yes, but he should be prepared for the lane to be closed farther down the road. 3) No, that lane is closed right there. 4) No, that symbol will change to red in a few seconds. cConfidence Question: How confident are you that your answer is correct? 1) Not at all confident, basically guessing. 2) 50% confident, I could narrow it down to two of the options. 3) 75% confident, pretty sure I’m right. 4) 95% confident, slight chance I’m wrong. 5) 100% confident. dTest Question 2 (Q2): Did the presence of the STALLED VEHICLE AHEAD message affect your answer? 1) No, I understood the arrows alone. 2) Yes, it helped explain the purpose of the arrows. eN = Sample Size fNA = Not Applicable Discussion e majority (84%) of the participants thought the supplemental ground-mounted signs explaining the symbols’ meaning were either helpful or very helpful. Regarding how frequently the signs should appear, the results are varied and inconclusive. Research Question 3: How does the presence of other CMS(s) displaying travel information on the same sign bridge as the lane-control signs affect comprehension and driving behavior? Overview Research Question 3 explored whether the addition of a CMS would aid in comprehension. All participants saw a sign bridge with two downward green arrows and two diagonal yellow arrows with approximately half of the participants having a supplemental CMS in the image (see Figure 35) and the other half not having the supplemental CMS. e CMS did not provide any information that explains what the symbols mean. Instead, the signs displayed information about the driving conditions. Because participants who saw Figure 35 may have had a com- prehension advantage on other questions, the test questions for Research Question 3 appeared toward the end of the testing session. Results Figure 49 provides the percentage of responses for each treatment tested within Research Question 3. It also provides the distribution of responses to the condence question along with the number of participants who saw that treatment. e Fisher’s exact tests conducted to deter- mine whether the distributions of responses for the treatments diered among the unpaired

64 Application of Dynamic Lane-Use Control: Proposed Practices data found that the distribution of responses for with and without the CMS was not statistically significant (p-value = 0.394). Discussion This question examined the effect of adding a CMS to the vertical support of the sign struc- ture to provide additional traffic information that could explain why a lane is closed or traffic is slowed. The results show a high level of comprehension for lane-use controls both with and without the explanatory CMS and found no statistical difference between the two conditions.