Communicating Changes in Horizontal Alignment (2006)

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.

2CHAPTER 2 FINDINGS LITERATURE REVIEW The literature review included examination of traditional sources and readily available reports. It should be noted from the outset that there appears to be a considerable body of lit- erature that is not documented in any traditional way and thus is not included. This includes work done by various agencies that is either not documented in any usable way or only avail- able within the agency. An example of the former would in- clude a TCD treatment that was implemented in the field but never documented or studied in any methodical way; an exam- ple of the latter would include an interoffice memo that might summarize an implementation and a brief study at a specific location. Such documents include unwritten guidelines for placement of various devices. For example, a local jurisdic- tion (in Michigan) has “rules” for when chevrons are used as well as for choosing the number and placement locations— these “rules” are not published but simply “known” within the agency. Such problems notwithstanding, some of the more inter- esting overarching results from the literature review are sum- marized immediately below and then followed by comments regarding specific types of devices: • The MUTCD provides only general guidance on the selection and application of TCDs used to inform drivers of a change in horizontal alignment. • Novelty effects should be carefully considered in any new TCD evaluation. • Currently, advisory speed signing appears to be largely ineffective if the goal is for drivers to actually travel at the posted advisory speed: drivers either fail to notice advisory speed plaques, or, more likely, they simply re- ject the literal advisory speed recommendations, driving at a reduced speed that they feel is appropriate. • Because raised pavement markers and post-mounted delineators provide both far and near guidance up to and through a curve, delineation should be part of a com- prehensive curve-risk reduction program. Also see more specific comments below. • Specific recommendations for curve TCDs on low- volume rural roads are not provided. • Benefits and guidance for using new TCDs in the MUTCD (e.g., the combined alignment/advisory speed sign) are not provided in the literature. Summary comments from the literature review pertaining to different types of devices follow. Curve and Turn Signs The literature review pertaining to curve and turn signs re- sulted in the following summary comments: • The messages conveyed by curve-related warning signs are likely more generic than traffic engineers might hope. Messages are probably weakened by driver limitations in perceiving and understanding nuanced curve/turn warning signs, which reaffirms the need for redundancy at highest-risk locations. • While there are varied results regarding whether curve- related signs reduce crashes, run-off-the-road (ROR) and single-vehicle crashes are, nonetheless, probably reduced when such signs are used (when compared with no signs). • If conventional curve-related signs are ineffective at high-crash locations, there is some evidence that special treatments such as oversized or traffic-actuated signs with beacons are effective. • The longitudinal placement for curve signs is typically based on the 85th-percentile or posted speed. This is con- sistent with the sign placement table in the MUTCD. Advisory Speeds The literature review pertaining to advisory speeds resulted in the following summary comments: • Some studies indicate that advisory speed plaques are no more effective than curve/turn signs alone. Conversely, others found that there was some speed reduction asso- ciated with the placement of advisories although not to the posted speed per se—that is, advisory speeds are rou- tinely exceeded. • Drivers may underestimate their actual speeds on curves. • Surveyed practitioners feel that guidelines for assigning advisories are sufficient. • Some results indicate that single-vehicle and ROR crashes are reduced when advisories are posted and that “sections” (i.e., a segment of road with numerous curves) with advisories posted experience fewer crashes.

Chevrons The literature review pertaining to chevrons resulted in the following summary comments: • Drivers have been noted to shift away from chevrons as they negotiate curves—toward the centerline on curves to the left, away from it on curves to the right. • Several studies noted that average (day and night) speeds increased when chevrons were added. • Crash reductions have been noted on curves marked with chevrons where standard curve-related signs have not been effective. • While chevrons provide additional guidance for the driver, perhaps not more than post-mounted delineators or raised pavement markers. • Chevrons have more effect at night, on sharper (>7°) curves, and when used in conjunction with edgelines. Edgelines and Centerlines The literature review pertaining to edge- and centerlines resulted in the following summary comments: • Results regarding increase or decrease in vehicle speeds when edge- and centerlines are used are varied: some studies show an increase, and others show a decrease. • Freshly painted lines tended to decrease lateral placement variance. • As above, edgelines seem to have more effect when used with chevrons. Post-Mounted Delineators The literature review pertaining to post-mounted delin- eators (PMDs) resulted in the following summary comments: • Similar to chevrons, drivers tend to shift away from PMDs—that is, away from the edge of road and toward the centerline. • Day- and nighttime speeds tend to increase when PMDs are used. • PMDs have been shown to reduce variance in lateral placement. • PMDs should only be placed on the outside of the curve and could be confusing if placed otherwise. • PMDs have more effect when used with freshly painted centerlines. Raised Pavement Markers The literature review pertaining to raised pavement markers (RPMs) resulted in the following summary comments: • RPMs are more effective at reducing crashes at high- crash locations than elsewhere although general reduction was not noted. 3 • RPMs tend to have more effect during the night than the day. • RPMs had more effect on lateral placement over and above freshly painted centerlines over time. • The variability of speed and lateral placement was de- creased with RPMs. • There was some evidence that RPMs had greater effect than PMDs and chevrons. It is also worth noting that the recent NCHRP Report 518: Safety Evaluation of Permanent Raised Pavement Markers contains a definitive review of the effectiveness of perma- nent RPMs. While some results were mixed and/or inconclu- sive, significant crash reductions were noted for wet-weather crashes and especially at night. Discussion and Comments Notwithstanding the impact of different TCDs in different circumstances, improvements to road geometry are clearly preferred to warning signs or delineation at locations with hazardous changes in horizontal curvature. However, where right-of-way, economic, or other practical limitations pro- hibit or delay desirable improvements, standard TCDs can be used effectively to warn drivers and to guide them through the potentially hazardous road segment. Identifying hazardous road segments is best accomplished by considering curve operational and physical properties and crash history. Posted, design, and operating speeds alone can- not predict overall safety on horizontal curves. It appears that aggregated crash rates are positively correlated with high degrees of curvature: curvature of approximately 5° or greater poses an increased hazard of which the driver should be warned. Other results showed that, in practice, the sharpest curves already have extensive TCD treatment and that some modest curves were among the most problematic in terms of higher crash rates and less treatment. TCDs used on horizontal curves should be applied uni- formly to foster desirable driver expectancies. In placing curve signs, engineers should consider preview sight distance, driver perception of road geometry, and drivers’ expectations for curve warning. In addition to curve warning signs, delin- eation will improve driver perception of alignment changes. Other roadside information is also an important information carrier—any practical measures that can be taken to improve a driver’s curve perception should take priority over warnings. While most of the TCDs for curves have been shown to be effective in certain experiments, the testing on which the results are based was rarely comprehensive or conclusive across the spectrum of conditions encountered in the field. For example, some work has shown that chevrons, PMDs, and RPMs result in higher speeds on curves. Given that most curve crashes occur because the motorist enters the curve at too high a speed, would we want to increase motorist speeds on all, or any, curves? Clearly, a definitive guideline is hard to develop based on the results of existing work.

A couple of general recommendations emerge from the lit- erature review. The first is the need to define more explicitly what constitutes an “engineering study” in the context of horizontal curves. Items that should likely be included in such a study for TCD selection and placement at horizontal curves include consideration of the following: • How other potentially limiting factors such as weather conditions would affect curve perception and vehicle performance; • Natural features of the road and roadside environment that provide delineation; • Alignment that drivers may find difficult to perceive or interpret; • Driver expectations; • Application of both the preview sight distance concept and 85th-percentile speed to longitudinal warning sign placement guidelines; • Degree of curvature and other parameters associated with high-risk alignments; • Guidelines for when redundant warning or warning along with delineation is recommended; and • Guidelines for when curve warning is not recommended based on curve or volume parameters apart from crash history. The second need is to recommend advisory speeds based upon driver comfort or an assessment of how fast drivers are already driving. Associating current operational speeds under ideal conditions with curve properties and ball-bank indicator ratings would suggest a more realistic and, perhaps, a more respected advisory speed recommendation. Implicit in the above is the need for consistent messages to be given to drivers. While drivers may not respond as well to different speed advisories as engineers would like insofar as drivers don’t slow down to the advisory level, it seems clear from prior studies that there is some response. Given the findings from the literature review, it was neces- sary to ascertain the current state of the practice relative to the specification and use of curve-related TCDs as well as deter- mining how drivers respond to them. For example, when and how are curve-related TCDs used and is that use consistent across different jurisdictions? Likewise, do drivers use the information from TCDs as intended and are the TCDs per- ceived to be used consistently across jurisdictions? The next sections document the tasks that were undertaken to determine the perceptions of practitioners and drivers with respect to the TCDs used for horizontal curves: are they ade- quate, are they used consistently, and should guidelines for use be changed? Practitioner input was captured through the use of focus groups, discussions in several states, and a nationwide survey. It should be noted from the outset that there were some problems throughout the project with specifying exactly which edition of the MUTCD is relevant. When this project 4 began, the 2003 edition of the MUTCD was not in wide use and, indeed, some practitioners weren’t even aware of the millennium edition. It still had not been officially adopted or used by all jurisdictions in mid-2005. So, somewhat arbitrar- ily, the millennium edition was used as the default “standard” throughout the project—e.g., as a reference point for practi- tioner survey instruments. Differences between the millen- nium and 2003 edition are addressed in later chapters. PRACTITIONER FOCUS GROUPS The practitioner focus group exercises were held in three states: Michigan, Indiana, and North Carolina. During the exercises, the practitioners who were responsible for specify- ing and implementing TCD treatments for horizontal curves— including both engineers and technicians—were brought together to discuss those treatments. The focus of the dis- cussion was on local, typically county-level, two-way, two- lane, rural roads. Topics included the adequacy of existing guidelines for the specification and placement of various TCDs, the adequacy of currently used TCDs, the identifica- tion of “problem” curves, the use of engineering judgment and studies, the appropriateness of advisory speeds and meth- odologies for setting them, and the reasons for perceived differences in treatments among jurisdictions and general inconsistency in the use of curve-related TCDs. In all, there were four formal focus-group exercises (the number of prac- titioners attending is shown in parentheses): two in Michigan (15, 9) and one each in Indiana (5) and North Carolina (10). In addition to the formal focus groups, there were also two separate interview sessions with practitioners in North Car- olina and one in Tennessee. The interview sessions included only one to three participants and therefore lacked the inter- active nature of the focus groups. Nonetheless, some useful information was gleaned from the sessions. The four focus-group exercises provided valuable infor- mation regarding the state of the practice for using curve- related TCDs. While all groups were familiar with and regu- larly used the MUTCD or a state-level variant as a guide for their curve treatments, some agencies supplemented them with additional materials that were not necessarily written. For example, there was a guideline in some Michigan coun- ties that indicated that chevrons should be used on any curve requiring a speed advisory. As a general statement, the MUTCD was perceived to provide sufficient guidance but leave room for judgment on the part of the practitioner. In this context, the standard curve signs (W1-1 through W1-8) were considered sufficient in most situations. The winding road (W1-5) and reverse curve or turn signs (W1-3, W1-4) cause some confusion in application, especially when advi- sory speed plates are to be used. The confusion results from the determination of which curve in a sequence should be the basis for the advisory—the first or the most severe? Beyond the standard curve-related signs, RPMs, chevrons, and large

target arrows appeared to be widely used as supplemental devices. Engineering judgment, as defined by the practitioners, is decisionmaking based on engineering or other technical train- ing, experience, and/or common sense although the latter is defined by the practitioners. While there was an implicit under- standing of engineering judgment and engineering studies, it was clear that not many studies were done to support the TCD-related judgments that were routinely rendered. In many instances, it appears that the day-to-day specification of TCDs for curves is left to technical support staff and not necessarily done by engineers except in atypical cases. To assess appropriate advisory speeds, most of the practi- tioners used a ball-bank indicator in some way. Based on more informal discussion with some of the practitioners, it is suspected that some practitioners still often use a “tried and true” method of simply driving the curve to assess the need for an advisory plate. Notwithstanding the assertion that ball- bank readings are typically used to assess appropriate advisory speeds, practitioners generally agreed that almost all curves signed with advisory speed plates can easily and safely be traversed at “+10” mph over the posted advisory speed with the possible exception of roads in mountainous areas. The practitioners also believe that drivers perceive the advisory speeds for curves in the same way: as a general guideline that says that one should slow down. It was not clear, however, if any of the practitioners had ever actually done a study at a curve to see whether deployed TCDs were effective in getting drivers to slow down or “comply” with the suggested advisory speed. At the same time, there were no advocates for changing the perception that posted advisory speeds were “too low”— in fact, the opposite view was typically expressed. All groups of practitioners agreed that there is some vari- ance in curve-related TCD selection and placement in their state although not within their own jurisdiction or other geo- graphic area for which they are responsible. Most typically, practitioners thought their own practice and/or practice within their jurisdiction was consistent but that “other jurisdictions” introduced inconsistency. The Michigan group indicated that this is partially due to the flexibility afforded by “engineer- ing judgment” but mostly because of the differences in bud- gets allocated to TCD deployments. The latter resulted in practitioners in some jurisdictions being “comprehensive” when it came to signing and marking horizontal curves. Others tried to get by with a bare minimum of TCDs. It was the general feeling that additional guidance in placing TCDs is needed, but it was fairly clear that the guidance that the MUTCD does offer was not necessarily being followed. There was also a concern noted that more clearly expressed guide- lines might lead to more litigation. This was interesting in Michigan given that there is currently a high degree of legal protection for TCD use by practitioners and jurisdictions. Given that in many county departments signing curves is a small part of a typical engineer’s job, charts or tables that would make the job easier and more consistent would be wel- 5 comed. At the same time, there are problems with such sim- ple guidelines unless the words “should” or “shall” are used. One Michigan group mentioned that such guidelines might also help to prevent over-signing of curves. Virtually all practitioners noted that horizontal curves are most dangerous to drivers in combination with other risk fac- tors such as driveways, vertical curves, geometric or design problems such as inappropriate superelevation, weather, and anything that leads to a violation of driver expectancy. As stated earlier, one of the objectives of this project was to identify the need for an improved methodology to deter- mine what information about horizontal curves is needed and how best to communicate it to drivers in a consistent and credible manner. From the practitioners in the focus groups and interviews, it is clear that inconsistency in use and de- ployment of curve-related TCDs is perceived to exist. How- ever, the reasons for this inconsistency were thought to be related to budgets, personnel turnover, the exercise of engi- neering judgment, and differences in topography between areas rather than differences in interpreting manuals or other rules or methodologies. Overall, there was a disconnect re- garding the need or desire for well-defined guidelines or rules—while it was allowed that more refined or explicit guidelines would lead to increased consistency, many, if not most, practitioners still valued flexibility and discretion based on “judgment,” and some objected to guidelines that could lead to liability issues if they were not followed. The practitioners in the focus groups believe that the worst kinds of curves are those where there is a combination of the curve with other risk factors, a condition that does not lend itself to a straightforward or necessarily uniform solution. These concerns are not easily converted into new or improved guidelines or, at least, not into very explicit ones. With regard to advisory speeds, most practitioners ap- peared to use the ball-bank indicator. While the “rules” that they used varied, there was consistency in the perception that advisory speeds were typically lower than they need to be. However, in the discussions, no alternative method for assign- ing the advisory speed received endorsement although some were opposed—for example, some practitioners thought that drivers already go too fast on curves and thus opposed using average or 85th-percentile speeds as being “biased” toward the high side. As a general rule, all practitioners indicated that in their view, drivers expect advisory speeds to be low and often exceed them. (This latter point was clearly not some- thing that the practitioners had either done studies on or col- lected any data to support their view.) Changing the method of determining the numeric value of the advisory speed to be more consistent may not be necessary since different methods yield the same results. Changing practice to be more credible would imply raising the advisory speed in most cases. This would actually lead to larger inconsistencies in the short term, and the risk associated with such a wholesale change in prac- tice and driver education may not be worth taking. Practi- tioners should be consulted once a standard methodology is

developed to gain their input on how well the procedure would work and the likelihood that those responsible for plac- ing curve signs will comply. PRACTITIONER SURVEY To obtain a broader view, a nationwide survey of practi- tioners responsible for specifying and implementing TCD treatments for horizontal curves was undertaken. Consistent with other project tasks, the emphasis was on rural two-way, two-lane roads. An estimated total of 1,250 practitioners from state, county, and city agencies in all 50 states were sent a 20-question survey by mail or e-mail; 344 responded. Those who received the survey were asked to respond and forward it to as many people as they deemed qualified to answer the questions. For this reason, an exact response rate cannot be determined. However, based on the 1,250 that were sent out, 344 responses would represent a response rate of ∼25%. Of the 344 practitioners that responded, 177 (52%) were from state agencies, 144 (42%) from county agencies, and 20 (6%) from city agencies. City officials were not targeted for re- sponses because of the project’s rural-road focus. In states where the state DOT has jurisdiction over all roads within the state, county responses were not available. The largest num- ber of responses (37, 11% of the sample) came from Michigan counties, which could influence the results. In most cases, however, the Michigan responses did not vary significantly from the others. In cases where the responses from Michigan were substantially different from the rest of the sample, both sets of responses were examined and reported. The practitioner survey provided some valuable informa- tion about the state of the practice for curve-related TCDs. Based on the relatively high level of response and the numer- ous comments made on survey forms, it is clear that the sign- ing and marking of horizontal curves is an issue that many practitioners find important. However, some responses were not easily interpreted. For example, one of the primary responses to the question “Which horizontal curves are the most difficult to sign and mark appropriately?” was “inter- section on curve.” However, there is a standard intersection- on-a-curve sign (W1-10) available in the MUTCD, so it would be expected that this situation would not be difficult to sign. It could be that respondents who checked this answer felt that this sign was not adequate for communicating the sit- uation to drivers, or they may not routinely use all TCDs at their disposal. Not surprisingly, agencies generally sign their own roads. When they do not, it is usually done by a higher government authority. While states and counties usually employ engineers and technicians, cities and townships often have other per- sonnel such as a sign manager or road foreman responsible for the selection of TCDs for horizontal curves. The large numbers of non-engineers involved in day-to-day signing calls to question when, how, and by whom “engineering judgment” 6 is being used in signing and marking curves. It may well be that routine decisions are made by technicians with engineers becoming involved only when atypical or more difficult sit- uations are encountered. However it is still the non-engineer who is defining such situations. The MUTCD is the most widely used guideline because most respondents claimed that it (or a similar state version) was used in their jurisdiction. The ball-bank indicator is the most common device for determining advisory speeds although many respondents use road geometry or test-drives in addition to or instead of the ball-bank indicator. Many respondents indicated that much of the training for sign spec- ification is done “in-house.” While standards and guidelines are accepted, they are not always followed. In-house and other forms of peer-to-peer training can lead to consistency over the years within a jurisdiction, but not necessarily among jurisdictions. This inconsistency among jurisdictions can also be seen in the low level of response to questions asked about practices in other jurisdictions and in the fact that respondents thought that practice such as setting advi- sory speeds was done less consistently in “other” jurisdic- tions relative to their own. Respondents also often answered “I don’t know” or left these questions blank, which also implies a potential lack of sharing of best practice among jurisdictions. Curves within the respondents’ jurisdictions were reported to usually be signed and marked according to the standards. Advisory speeds seem to be determined somewhat consis- tently within jurisdictions but not among jurisdictions or across a state. The potential differences in procedures for determining advisory speeds notwithstanding, the consensus is that many, if not most, speed advisories are lower than needed and, most assuredly, not too high. Questions about consistency in the use of curve signs and in particular advisory speed signs yielded lower response rates. This may be due to the fact that respondents are worried about liability or to the fact that they are not confident that all the curves in their area have appro- priate signs or advisory speeds. Respondents did not feel that curve TCDs were overused, and some even commented that signs might be underused in places. One common observa- tion was that budget and staff limitations have an effect on the selection and placement of TCDs on horizontal curves. Many respondents pointed out that personnel in many more rural jurisdictions know the standards, but do not have the financial resources, the staff, or the equipment to perform studies and/or place signs. Similar to the focus group results, the clear picture that emerged was that rural jurisdictions and those with limited budgets have less comprehensive curve- related signing than more urban or richer jurisdictions. When broken down into state and county jurisdictions, answers about signing consistency did not change greatly. State-level respondents tended to be more uniform and posi- tive in their answers than were their county-level counterparts, but still had a negative view about practices across the state or at different jurisdictional levels within a state. At the state level,

curve TCDs are reported to be used more often than at the county level. This is probably due to the fact that state roads tend to have higher traffic volumes and are generally held to a “higher standard” than are county roads and to the fact that state road departments have more resources available to them. With regard to identifying curves that are difficult to sign or cause problems for average drivers, there was some dis- agreement. State and county-level respondents tended to agree that the five most-difficult curves to sign are “broken- back” curves, curves that get sharper as the driver goes through them, spiral curves, curves over the top of a hill, and back-to- back curves. However, they indicated that the five types of curves that are the most-difficult to drive are spiral curves, curves over the top of a hill, unexpected curves, sharp curves after a long tangent, and very sharp curves. According to the respondents, curves that are most difficult for drivers to nego- tiate are those that violate their expectations, that include a combination of factors, or both. In conclusion, the state of the practice with respect to curve- related TCDs on rural two-way, two-lane roads is consistent in some aspects, but not in others. Both the focus groups and the survey yielded similar results. Standards and methods cited are relatively uniform, and variation in the standard infor- mation that is used appears to be minor. From what the prac- titioners have stated, the standards are generally understood and followed in their jurisdiction, although “in-house” train- ing or practice sometimes overrides printed guidelines. Most agencies feel that they have achieved consistency within their own jurisdiction, but are less positive about what is hap- pening elsewhere. State-level agencies were more confident than were county-level agencies that curve signing and advi- sory speed determination practices were the same across their jurisdiction. The biggest potential problem area appears to be in assessing and establishing advisory speeds. As a point of interest, if the 2003 version of the MUTCD is followed, it can be expected that there will be even more inconsistency among jurisdictions and that drivers will see some curves undergo significant changes in the advisory speeds that are posted. The practitioners surveyed in this study seemed to all be of the same mind with respect to which curves are most difficult to sign and drive. Better communication among agencies, either through state oversight or technology transfer programs, may be needed to improve consistency. DRIVER FOCUS GROUPS The perceptions of drivers regarding the treatment of hori- zontal curves were also assessed. While information from drivers is useful in its own right, it is also useful to compare their perceptions with the practitioners’. In addition to assess- ing driver perceptions, an attempt was made to differentiate between those who had been involved in a crash on a hori- zontal curve and those who had not, the idea being that a com- parison between curve-crash-involved and “typical” drivers 7 might reveal different patterns in driving habits or in how they respond either to TCDs or to the curves themselves. The first task was to convene focus groups of crash- involved and typical drivers. Two separate focus groups were held in East Lansing, Michigan, in March and June 2004. The intent of the focus-group exercises was to discuss TCDs used on rural, two-way, two-lane horizontal curves. These sessions were spent discussing TCDs in general, iden- tifying actual problem curve situations encountered by the participants, and then reviewing the TCDs that had been deployed and how they might be modified to make the situ- ations less problematic. There was also discussion about re- sponses to advisory speed signs and the general topic of TCDs for horizontal curves. As noted above, the initial goal had been to convene one focus group of around 10 drivers evenly split between those who had been involved in crashes on rural curves and “typi- cal” or average drivers who had not been involved in that sort of crash. Recruitment for the first exercise did not go accord- ing to plan, and a smaller group was used (five participants), only one of whom was known to have crashed. While the potential crash-involved participants were identified using crash records, the relatively limited numbers made recruit- ment difficult. The non-crash participants were simply iden- tified using local and web-based telephone directories. As it turned out, the smaller group was better for the type of activ- ities undertaken and allowed ample opportunity for discus- sion and interaction by all participants. A second group of six participants was convened later although it contained no pre-identified crash-involved participants. Participant ages were estimated to range from the early 20s to mid-70s with five male and six female participants. All participants were from suburban or rural areas in the three counties surround- ing the urban area of Lansing/East Lansing, Michigan. The telephone recruiters who contacted potential participants concentrated on exchanges from suburban and rural areas so that there would be a good experiential base with rural roads and curves. All participants had experience with both urban and rural streets and roads in the area. In general the participants in both focus groups were knowledgeable about the typical TCDs used for rural curves. They indicated that they have lower expectations for the num- ber of TCDs present on low-volume facilities. At some point, both groups indicated that there needs to be better mainte- nance of TCDs on rural curves, whether it is by re-striping edge lines or clearing brush around signs. In addition, the participants commented that there is a need for better con- sistency among states and, more specifically, among counties in Michigan. In terms of the specific TCDs that participants think would be helpful at curves, participants indicated their preference for • Advance curve warning signs with advisory speed plates; • Information (via signs) regarding other potential issues at the curve such as an intersection on the curve or loose gravel; and

• Seeing the curve in advance, especially through the addi- tion of reflective visual aids for nighttime driving such as chevrons or by clearing obstruction near TCDs. Participants indicated that they tended to drive slower and closer to the posted advisory speed when driving unfamiliar curves. When driving familiar curves, participants indicated that they routinely go at least 5 mph over the advisory and/or about 5 mph faster than they would at an unfamiliar curve (the responses were not necessarily consistent). In general, speed advisories were perceived as a warning to “slow down” but not necessarily to the advisory level. In that context, there was considerable variance in how speed advisories should be used and at what level they should be set. However, the participants indicated that they felt that changing advisory speeds at rural curves is not a high priority for change or use of financial resources. Most of the participants have their own method of relating their speed to the posted advisory speed. CRASH-INVOLVED AND TYPICAL/ AVERAGE DRIVER SURVEY As noted, both a focus-group exercise and a survey were undertaken to assess driver responses to different curve- related issues. The focus of the survey was also on rural two- lane, two-way roads. The topics were parallel to both those of the driver focus group and the practitioner-related exercises and included the adequacy of existing signs; perception of and response to specific TCDs, identification of what makes some curves more difficult than others, the extent to which famil- iarity is a factor in difficulty with curves, what types of roads present the most difficulty, how TCDs for curves could be improved, characteristics of their own “worst” curve-related experience, and whether that “worst” experience on a curve had short- or long-term effects on driving habits. The drivers were broken into two basic groups: drivers involved in a motor- vehicle crash on a horizontal curve within approximately two years; and typical or “average” drivers. The survey was done in both Michigan and North Carolina. There were actually seven different driver groups queried: typical Michigan-local, typical Michigan-statewide, typical North Carolina, two sub-groups of crash-involved Michigan-local, crash-involved Michigan-statewide, and crash-involved North Carolina. The differentiation between local and statewide Michigan drivers occurred because there was interest in identifying a “local” crash group for other phases of the project. The local group was composed of drivers in and around the greater Lansing area, which included both the urbanized Lansing/East Lansing area as well as the surrounding six counties that are largely rural agricultural areas outside the immediate urbanized area. The Michigan statewide sample was from the Lower Peninsula only. The overall response rate was ∼13%. The following summary and discussion are based on stated preferences and do not necessarily represent how respon- dents actually perform on the road system. The summary is 8 provided in three separate categories: where there were no or slight differences among the respondent groups, where there were differences by state, and where there were differences by crash versus typical drivers. The first set of summary statements contains those where there were no substantial differences among the groups of respondents: • Except of the local Michigan crash group, respondents were generally confident that they knew the difference between advisory and regulatory signs. • Respondents generally stated that advance curve warning signs helped them anticipate and drive through a curve. • Respondents generally stated that large arrows and chevrons as well as pavement markings helped them anticipate and drive through a curve. • The most prevalent comment about advisory speeds was that their use “depends” on one or more factors such as weather and visibility. Many respondents commented that they generally interpreted the advisory speed as a message to “slow down,” even if not to the actual speed noted on the advisory sign. Few respondents reported actually driving at the advisory speed; conversely, numer- ous comments were made to the effect that advisory speeds are routinely exceeded. • There was no consensus on how appropriate advisory speeds should be determined: fully 25% thought it should be the maximum safe speed, more than 30% thought it should be faster than most drove, and 36% thought it should be a “comfortable” speed on dry pavement. • While speed was often mentioned in the context of prob- lems with curves, 42% of all the respondents indicated that their worst-curve experiences occurred on roads with speed limits of 50 mph or less. • The “top five” characteristics of problem curves include very sharp, no advance signs, over top of a hill, getting sharper as the curve is traversed, and unexpected. With minor variations, these were listed by all groups. Violation of driver expectation appears to be the common thread. • The “top three” sign-related solutions were to lengthen the distance between the advance signs and the curves, to increase sign size, and to increase the number of advance signs. All of these are concerned with giving the driver the appropriate information earlier and appear to be consistent with the “top five” problems. • In the context of fixing problem curves, while TCD changes were often mentioned as remedies, other fac- tors such as geometric changes were mentioned more often. This is in spite of the clear emphasis in the survey instrument on TCDs. • If a curve “check list” was to be made on how to “fix” curves, it would include correct the superelevation; install/ maintain good-quality center- and edgelines; ensure that appropriate advance warning signs are present; generally make sure that the road, shoulders, and right-of-way are

well-maintained; and provide special emphasis on “bad” or dangerous curves. In some instances, there were variations in answers to ques- tions that could be attributed to the respondent’s state of resi- dence. While these are highlighted below, it is not clear that the variation is really state- or region-related. In many instances, the variations among the various groups in Michigan were more pronounced than those between the two states. Although the majority of all respondents indicated that they used advisory speeds as “guides,” Michigan respondents were more likely to use them in that manner and less likely to say that they drove at the advisory speed. It is not clear why this difference exists—it may be a function of the perception of speed-limit enforcement in general. The variations in answers that could be attributed to re- spondents’ state of residency are as follows: • While all respondents thought familiarity was a signifi- cant issue in successfully negotiating curves, it was rated as more important by Michigan respondents. • Michigan respondents were more likely to report gravel roads as a contributing factor in making curves haz- ardous. This may simply be a function of there being more gravel roads in rural Michigan. • The North Carolina respondents were more likely to cite a combination of factors in describing their worst curve experiences. Finally, there were differences between respondents who had been involved in a curve-related crash and the “typical” respondents. While some crash respondents may not have been involved in crashes on curves because of difficulties in locating actual crash locations, they had certainly been in- volved in some crash; likewise, while some of the typical, randomly selected respondents may have been involved in crashes, it should be much less likely. These caveats notwith- standing, differences between the two groups included the following: • Crash groups in both states were more likely than their typical counterparts to disagree with a statement indi- cating that they were getting enough advance informa- tion about upcoming curves. • Although all respondents were generally positive about advisory speed signs, crash groups were more likely to disagree with a statement indicating that advisory speed signs were helpful. • While crash-involved respondents were not overwhelm- ingly honest in reporting their “worst curve” experience, they were far more likely to note involvement in a crash than were typical groups. • Typical respondents were more likely to think that curves on gravel roads were a more serious problem than did the crash respondents. 9 • Typical respondents were more likely to think that curves “far away” from home were the most problematic while crash respondents indicated that “closer” curves were—that is, it is likely that those who crash on curves do so more often on roads with which they are familiar. This is in spite of an earlier assertion that familiarity with curves was a significant issue for all groups. • While all respondents cited poor weather or pavement conditions, poor visibility, darkness, and unexpected traffic as important contributors to curve problems, the crash groups cited poor weather more often. Conversely, darkness was more often cited by typical drivers. • While all groups cited speed-related problems when de- scribing their “worst curve” experience, the crash groups were less likely than their counterparts to cite “going too fast” as part of the problem. Ironically, most groups thought that advisory speeds are too low while citing going “too fast” as a contributing factor in worst-curve experiences. • Crash groups were somewhat more desirous of change in signs and markings. Overall, approximately 25% of the respondents did not think any TCD-related change was necessary. • Crash groups were somewhat more likely to think that bad curve experiences had an effect on driving behavior although all groups stated that, not surprisingly, the effects lessened over time. Overall, while there were some differences in responses by state, it is not clear that the differences are related to intrinsic regional behavior differences or simply differences in the roadway system such as fewer gravel roads in North Carolina. On the other hand, there were often clear differ- ences between crash and typical respondents. Drivers who had been involved in crashes were more likely to state that their worst experiences occurred closer to home than were the typ- ical respondents, and they were more likely to indicate the need for more or better communication. While the former confirms what is known about many crashes occurring closer to home, it is not clear what the implications are for TCDs. Improved communication would probably not help drivers in many of their familiar problem situations. On the other hand, all respondent groups stated that “fixes” that deal with geom- etry and other design features are more important than mod- ifications of TCDs. In regard to changes needed for TCDs, for the most part, respondents seemed satisfied with the ones they routinely encounter. However, they were adamant in sug- gesting that center- and edgelines be more widely used. Inter- estingly, there were very few comments about the use of chevrons, large arrows, or both. There is cause for concern with respect to speed advisory signs. While most respondents report using the advisories as a general guide and as a suggestion to “slow down,” they also indicate that they typically exceed advisory speeds. At the same time, they often cite excessive speed as a reason for

their problems with curves. There was also significant vari- ance in how respondents think that speed advisories should be set, from lower than most would go (which is the case now) to a maximum safe speed on dry pavement. FIELD STUDY OF DRIVER BEHAVIOR USING DRIVER PERFORMANCE MONITORING TECHNIQUE The project also included a driver observation study using driver performance monitoring (DPM) techniques in which a sample of randomly selected drivers were observed as they traversed an approximately 25-mile predetermined route negotiating 43 curves. As part of the DPM data collec- tion, detailed observations were made by trained observers at 11 curve sequences (some sequences contained more than one curve) and included assessments of a driver’s “search, speed, and direction control” performance as they negoti- ated each sequence. The vehicle’s speedometer readings at various points were also recorded as were comments on driving behavior. As applied in this project, DPM is an observation tech- nique in which one or more trained observers ride with a sub- ject driver (in the subject’s vehicle) over a predetermined route and make observations regarding driving behavior at specific locations. The behavior of each driver is compared with what is expected at each location. That expectation is established by observing other drivers at the same locations on the same route prior to the data collection runs. While DPM is a qualitative approach to evaluating driver behavior, it is made as quantitative as possible by scoring each subject as exhibiting satisfactory or unsatisfactory behavior on each of three dimensions: visual search, speed, and direction con- trol. An example of satisfactory behavior would be gradually slowing down for a curve versus an abrupt speed change right at the point of curvature. The observer(s) also makes comments regarding any other occurrences such as a vehicle pulling out of a driveway on the approach to a curve or other actions that might affect driver performance. The person in the front seat is the principal observer making the judgments on behavior while the back-seat observer makes ancillary comments and observations such as speedometer readings at various points along the route. DPM has been used in several contexts including as a driver-training tool for commercial truck drivers, as a diagnostic tool to evaluate whether recov- ering stroke victims are safe to resume driving, and as a research tool in an earlier NCHRP project on older drivers (NCHRP Project 3-44, “Improved Traffic Control Device Design and Placement to Aid the Older Driver”). General DPM Route Description The DPM route used in this project was an approximately 25-mile loop in a primarily rural area southeast of Lansing, 10 Michigan. The two-way, two-lane roads that composed the route were a mixture of a state-numbered highway and county roads, although the latter predominated. There were 43 curves on the route with 11 curves or curve combinations compos- ing the DPM observation sequences. The terrain is generally gently rolling with a few steeper sections. There were no long grades. The environment ranged from a low-density suburban area with houses set back 75 feet or more from the traveled way to wooded and agricultural areas with farms and widely scattered houses. The aerials of the DPM curve sequences also provide detail of the environment adjacent to the route. The prevailing traffic volumes were typical of rural roads, generally on the order of 1 to 2,000 vehicles per day. The state-numbered and -maintained road tended to be in better condition and to have better and slightly wider shoulders than did the county roads. Regulatory speed limits ranged from 45 to 55 mph although some were not marked. For the latter, the de facto limit in Michigan in rural areas is 55 mph. The curves themselves varied from very gentle with excellent vis- ibility, requiring no speed reduction or advisory speed signs, to a sharply descending right-angle turn, just after the crest of a hill, which had a 20 mph advisory speed plate. There was one ∼2-mile section of gravel road and another section that was a quarter-mile to one-half-mile long. Most paved sections had marked center- and edgelines. DPM Subjects The 39 DPM subjects were randomly selected from online and printed telephone books. The subject pool was con- strained by telephone exchange so that they would be from the area and, hence, more likely to participate. However, there was no attempt to randomize by age or to adhere to a rigor- ous statistical sampling regime as the purpose was not to achieve statistically significant results. The 39 subjects were split between day (28) and night (11) conditions. In addition to the randomly identified subjects, there were four Michigan State Police (MSP) officers who drove the route and were observed as if they were regular subjects. The MSP officers were instructed to drive at the edge of what they perceived as the “safe envelope” so that an approximation of the maxi- mum safe speed could be obtained. In spite of this direction, one of the officers drove at the posted regulatory and advi- sory speeds although he indicated that the “safe” speed was higher than those at which he drove—the data for this person were eliminated from all analyses. DPM Results As noted above, there are a number of quantifiable driver behaviors derived from the observers’ comments and obser- vations. For example, each driver either performed satisfac- torily or not on each of the standard DPM measurements of search, speed, and direction control through any given curve

sequence. Note that a sequence might consist of as many as three or four sections. Likewise, observer summaries con- tained indications of the type of errors, if any, that each driver made as he or she traversed the sequence, although these were not necessarily limited to the standard DPM measurements. It is important to note that the words “error” and “problem,” as used herein, refer to a deviation from expected driver behav- ior. Declaration of an “error” or a driver having a “problem” does not necessarily imply that a patently unsafe maneuver has occurred. Errors included not looking in the appropriate direction for an oncoming or intersecting vehicle, not slowing down appropriately, an abrupt and/or late speed reduction, en- croaching on or crossing the center- or edgelines or both in the case of a multiple-curve sequence, and handling difficulties such as abrupt or potentially dangerous lane changing or encroachment. While not all portions of the route were marked with edge- or centerlines, a centerline crossing was nonethe- less noted when the driver crossed the approximate center of the roadway. The DPM performance assessment and other comments were quantified and displayed in a spreadsheet. In addition, the speeds collected by the back-seat observers were used to construct an approximate speed profile for the entire route. Finally, the performances of all drivers on each DPM sequence were aggregated to provide overall “scores” for the sequences. Given this performance-based information, the sequences could be ordered in any one of several ways. For example, the sequences with the highest numbers of driver errors could be 11 identified or they could be ordered by the frequency of a spe- cific type of error. Likewise, the sequences could be ordered by approach speed or speed in the curve. The summaries of the subject performances can also be compared with the per- formance of the MSP officers or with other expectations of performance. An example of the former is a comparison of the difference between the subjects’ average speed for a specific curve in a DPM sequence with the average (n=3) from the MSP officers. An example of the latter is comparison between average subject speed and the posted regulatory and advisory speeds. Another basis for comparison of driver performance is a hierarchical ranking of “curve difficulty” or complexity based on the TCDs currently deployed. This hierarchy was devel- oped by project personnel and is shown in Table 1. Each curve on the DPM route can be assigned a number based on this hierarchy. Note that as used here, the hierarchy is based on the TCDs currently deployed at any given curve and not, for example, on any measure of the curve itself such as degree of curvature or radius. After all the DPM observations had been made, short subject-based reports were written and comments were aggre- gated for each sequence. Discussion was organized on the basis of the hierarchy of curve-related TCDs deployed— generally from the simplest curve to the most complex, based on existing signing. The hierarchy is subjective in nature but does provide a basis for ordering the discussion. The hierarchy is not without problems. For example, a curve Curve Ranking Signing Conditions 0 no curve 1 curve present, no sign 2 curve w/curve sign only 3 curve w/turn sign only 4 curve w/reverse turn sign only 5 curve sign + large arrow 6 turn sign + large arrow 7 reverse curve sign + large arrow 8 curve sign w/speed advisory 9 turn sign w/speed advisory 10 reverse curve sign w/speed advisory 11 curve sign + chevrons 12 turn sign + chevrons 13 reverse curve sign + chevrons 14 curve sign w/speed advisory + large arrow 15 turn sign w/speed advisory + large arrow 16 reverse curve sign w/speed advisory + large arrow 17 curve sign w/speed advisory + chevrons 18 turn sign w/speed advisory + chevrons 19 reverse curve sign w/speed advisory + chevrons 20 curve sign + other combination 21 turn sign + other combination 22 reverse curve sign + other combination TABLE 1 Curve hierarchy based on TCDs deployed

with a curve arrow and a speed advisory plate of 45 mph is “higher ranked” than a curve with a turn arrow but no advi- sory speed plate. By way of context, the values of the curve hierarchy for the DPM sequences ranged from 1—a curve with no curve-related treatments—to 17—a complex curve with curve warning signs, advisory speed plaques, and chevrons. In some instances, there are multiple curves in a DPM sequence although they are not necessarily reverse curves. The individual curves in a single DPM sequence may, and frequently do, have differ- ent values within the hierarchy. Finally, the sequences con- tain situations in which the worst curve was the first one encountered and other situations in which the worst curve was the second or third one encountered. In terms of obser- vations, there were 42 drivers including the 3 MSP officers. The number of drivers having one or more errors at the var- ious sequences ranges from 1 to 35. With one exception, all curve sequences presented problems to almost a quarter of all drivers; the “worst” sequences resulted in at least one error assigned to almost 90% of the regular (non-MSP) drivers. Discussion of DPM Results Given the DPM results for individual sequences, the ques- tion is whether there are overarching issues that emerge and, especially, whether there is anything that may impact the guidelines for the use of curve signs and markings. In the dis- cussion that follows, reference is made to specific DPM or curve sequences. Advisory Speeds Perhaps the most interesting finding that emerges from the sequence-based results is the response of drivers to advisory speed signs and to regulatory speed limits. First, there were three different speed limits encountered by drivers on the DPM route: 45, 50, and 55 mph. In addition, there were two gravel-road sequences where the speed limit was unmarked and, thus, a de facto 55-mph limit was in effect (Sequences 4 and 6). Signed advisory speeds for the DPM sequences ranged from 20 to 40 mph. Speed advisories were not present at all curve sequences, and none was present on either sequence on the gravel roads. On the curve sequences, the average speeds of the regular drivers never exceeded the posted or de facto speed limits. On the gravel road segments, regular drivers were quite conservative, approaching curves at average speeds less than 35 mph. The MSP officers drove similarly, although faster, in all instances except on the curve sequences on gravel roads where they drove about 2 mph slower than the other drivers. Considering those sequences marked with advisory speeds, in every instance drivers drove, on average, faster than the speed advisories that were displayed. Moreover, there is some indication that lower speed advisories did not result in 12 proportionately lower speeds. For example, for the two sequences with the lowest speed advisories of 20 and 25 mph, the daytime average speeds were 10 to 11 mph over the advi- sory, nighttime averages were 7 to 8 mph over, and the MSP averages were 12 and 16 mph over. For the four sequences with 35-mph speed advisories, daytime averages were 5 to 8 mph over, nighttime averages were 5 to 7 mph over, and the MSP averages were 11 to 15 mph over. For the sequence with the 40-mph advisory, the daytime average was 1 mph over, nighttime was at the advisory, and the MSP average was 7 mph over. For Sequence 1 where the second curve in the sequence had an advisory of 40 mph, daytime drivers were ∼6 mph over, nighttime drivers about 3 mph over, and MSP officers 11 mph over. Not considering the sequences on gravel roads, for the sequences with no advisory speeds and regulatory limit 50 or 55 mph, the daytime average ranged from 47 to 52 mph, nighttime from 44 to 48 mph, and the MSP from 52 to 57 mph. Based on these results, it is noted that while drivers slow down when confronted with speed advisories and slow down even more for lower advisories, they are still more likely to exceed lower advisories by a greater amount than they do higher advisories—their speed-reduction response is some- what limited. There are several possible explanations for this: drivers may feel that the risk of exceeding the advisory speed is lower at lower speeds; they may simply feel that they don’t need to go that slowly; or they may perceive that they have slowed down the requisite amount—that is, their sense of speed may be diminished. The latter explanation is similar to the reasoning for freeway off-ramp crashes where the appro- priate exit speed, sometimes as low as 25 mph, is misjudged in comparison with the mainline freeway speed, which can be 70 mph or higher—drivers are simply not going as slow as they think they are. Assuming that speed advisories are marked using the MUTCD as a general guide, drivers are consistently driving faster than the curves are marked. Moreover, the MSP speeds indicate that for most situations, the maximum safe speed is even higher. However, for curves with the lowest advisory speeds, the number of drivers with curve-related errors is the greatest over all sequences (25 and 24 of 39 regular subjects, respectively). Based on the results just presented, the question arises of what would happen if the existing guidelines for speed advi- sories were to be changed, resulting in advisories being in- creased. While there are different scenarios that might occur, the worst-case is that drivers maintain the difference between posted advisories and their travel speed. Under this scenario, if the curves with 20- and 25-mph advisories on the DPM route were increased 5 or 10 mph and drivers continued to exceed these lowest advisories by 10 to 11 mph in the day- time, they would be negotiating the curves at 35 and 40 mph or similar to the speeds of the MSP officers. An increase in curve-related driver errors could be expected. Under this sce- nario, where existing advisories were 35 mph and drivers

were 5 to 8 mph above the advisory, if these advisories were raised to 40 mph, drivers could be expected to drive at 45 to 48 mph—only a little under the MSP speeds of 46 to 50 mph. It remains to be seen how drivers would respond to increas- ing advisory speeds, but the evidence here suggests that there could be problems, especially at those curves requiring the lowest advisories. The findings with respect to drivers exceeding speed advi- sories is consistent with findings from other parts of this proj- ect where drivers reported that they consistently exceed posted advisory speeds at curves and where practitioners indicated that their expectations and perception of common practice were that drivers routinely exceeded advisories. In terms of recommendations, what emerges from this analysis, at least in the short term, is that raising advisory speeds may well result in problems for drivers who routinely exceed advisories. Moreover, the problem could be worse on curves with the lowest advisories. Thus, the effects of driver response to raised advisories should be thoroughly studied before changes are made. Given the numbers and types of driver errors with existing treatments, it seems appropriate that signing for curves with the greatest difference between the posted speed limit and the advisory speed should be com- prehensive (e.g., advance warning signs, speed advisories, chevrons, or centerline markings). While the use of compre- hensive treatments did not result in lower numbers of driver errors in the study, drivers should still be given as much in- formation as possible on curves where the desired speed re- duction is the greatest. General Speed-Related Issues Reordering the DPM sequences by the entry speed (which is defined as the lowest of the posted speed limit and the posted advisory) for the sequence and then examining the number of drivers with curve-related errors revealed an inter- esting pattern. The two sequences (9 and 5) with the lowest advisories (20 and 25 mph) had the largest numbers of drivers (24 and 25) having errors. The four sequences (2, 3, 7, and 8) with 35-mph advisories were next with 19, 12, 8, and 7 drivers having errors; and the 40- and 50-mph sequences (5a and 1) had even lower numbers of drivers with errors (1 and 4). How- ever, the sequences with the highest speed limits (55 mph) were mixed with three (6, 11, and 10) having a higher num- ber of errors (19, 18, and 13) and one (4) with only one driver making an error. Sequences 10 and 11 both consist of multi- ple curves and are ranked “high” because in each sequence, the initial curve has no advisory but there are advisories and additional treatments for a subsequent curve. Notwithstand- ing that Sequences 10 and 11 have advisories within the sequence, the preponderance of drivers making errors come at “both ends” of the speed scale—at curves that have very low advisories and those where the initial speed is high. While some issues could be taken with the definition of an error or 13 “problem” within the context of DPM, it seems clear that the problematic sequences that emerge are those that require sig- nificant speed reduction at the outset and those where the necessary reduction, although less, is within a sequence of curves. The recommendation for the sequences where the initial speed is relatively high but decreases at a subsequent curve within a series of curves is to continue to sign the lower- speed curve aggressively in spite of the number of drivers making errors even when speed advisories and other treat- ments are present. This recommendation is based more on common sense and judgment than the study results per se. Driver Errors The sequences were also examined by ordering them according to the number of drivers having errors. In addition to speed-related errors at the curves with the lowest advi- sories, the greatest number of drivers (24 and 25) making non-intersection errors occurred for these same curves. The next highest numbers of drivers having non-intersection errors occurred for a curve with a 35-mph advisory (19) and two that had no advisory (18 and 19), one of which had a posted limit and one of which was not posted. Thus, looking at the speeds and the driver errors, for at least some curves what happens is that speeds in excess of the advisories, and especially the lower advisories, lead to greater numbers of driver errors. Examining the driver errors in more detail, an arbitrary distinction between sequences where 10 or more drivers had curve-related errors versus those sequences where there were fewer than 10 is made. This shows that seven sequences fall into the first category while five fall into the second. Inter- estingly, the curves where more drivers made errors were also higher-ranked in the curve hierarchy—that is, these curves already had more extensive treatments. This was especially true with the use of chevrons where four of the seven in the first category had chevrons present versus only on one of the other group (of five). Contrarily, neither of the two worst curves had chevrons deployed. This leads to one of two conclusions: either chevrons have little impact in reducing drivers’ curve-related errors or drivers would have experienced even more errors if the chevrons had not been present. Examining the initial speeds for both groups of curve sequences reveals that while the average drivers approached the first category of curves a little faster than the other (∼41 mph versus ∼39 mph), the MSP officers showed the opposite (∼44 mph versus ∼46 mph). Interestingly, the two sequences on gravel roads showed opposite results. For one sequence (4), there were very few errors as drivers approached and drove through the curve con- servatively, negotiating them with ease. The other sequence (6), which consisted of a much more abrupt curve with less

visibility through the curve, was more problematic. The con- clusion, albeit based on only two sequences, is that drivers will generally perform well unless a curve is more difficult. This gives credibility to the idea that gravel roads may require less signing except at difficult curves such as those that require a turn arrow, have sight distance problems, or both. Again, the sample size of curves in this category is quite small. Curves and Intersections Results from other parts of this project have also indicated that more problematic types of curves are those that have additional distractions or elements—most specifically, an intersection either on or near the curve. The speed differen- tials notwithstanding, the problems that drivers had in sev- eral sequences were in not giving adequate attention to the intersection that was encountered in the vicinity of the curve in spite of the fact that standard intersection signs were typ- ically present. There were six sequences (5, 11, 3, 7, 8, and 5a) that had an intersection or, in one instance, a driveway that was in essence an intersection. The two sequences that had the most intersection-related errors were 5 and 5a where the intersection was literally on the curve: 10 and 25 drivers, respectively, made errors. In these two sequences the inter- secting roadway came in at close to a right angle from one side or the other. Three other sequences (11, 3, and 8) had intersections close to, but not on, the curve. These intersec- tions were far less problematic with between one and four drivers making errors. The final sequence of interest (7) had an intersection in the curve, but drivers had an excellent view of it as the intersecting road was essentially “straight ahead” as they approached it and the road (and route) curved to the left—a significantly different “look” for the approaching driver. Notwithstanding the fact that all of these intersections were marked with Intersection Ahead signs, it is clear that intersections in the curve caused the most difficulties for drivers. None of the signs “combined” the curve and inter- section warning. Given the problems experienced, even with signs deployed, it seems clear that the combination is indeed problematic and that drivers should be provided with advance warning whenever possible. Overall Summary and Recommendations Based on the discussion above, overall findings are re- iterated and some recommendations made. Advisory Speeds Findings and recommendations regarding advisory speeds are as follows: • The DPM subjects routinely exceeded posted advisory speeds. These findings are consistent with findings from 14 other parts of this project where surveyed drivers re- ported that they consistently exceed posted advisory speeds at curves and where practitioners indicated that their expectations and perception of common practice were that drivers routinely exceeded advisories. This phenomenon is also noted in the literature review. • Raising advisory speeds may well result in drivers who routinely exceed advisories and approach the maximum safe speeds, at least in the short term. Moreover, the problem will likely be worse on curves with the lowest advisories. Thus, the effects of driver response to raised advisories should be thoroughly studied before changes are made. • It is not clear that comprehensive signing for curves needing the lowest speed advisories will reduce the number of drivers with errors. However, the number of problems with such curves suggests that drivers either need additional information or need to better process and/or respond to the information that is made available. Thus, it is argued that signing should be comprehensive for the situations where the lowest advisory speeds are appropriate. General Speed-Related Issues A recommendation regarding speed-related issues is as follows: • The recommendation for the sequences where the initial speed is relatively high but decreases at a subsequent curve within a series of relatively close curves is to con- tinue to sign the lower-speed curve aggressively in spite of the number of drivers making errors even when speed advisories and other treatments are present. This is con- sistent with current practice of independent signing for sequential curves unless the tangent is <600 feet long. Driver Errors Findings and recommendations regarding driver errors are as follows: • When sequences were divided into two groups (≥10 drivers made errors versus <10), seven fall into the first category while five fall into the second. The curves where more drivers made errors already had more exten- sive treatments. This was especially true with the use of chevrons where four of the seven more problematic sequences had chevrons present versus only on one of the other group (of five). Contrarily, neither of the two worst curves had chevrons deployed. • The two sequences on gravel roads showed opposite results. For one sequence, there were very few errors as drivers approached and drove through the curves conser-

vatively, negotiating them with ease. The other sequence was more problematic with a much more abrupt curve with lower visibility through it. The conclusion, albeit based on only two sequences, is that drivers will gener- ally “take it easy” on gravel roads and will perform well unless a curve is more difficult or unexpected. This gives credence to the idea that gravel roads may require less signing except for “difficult” curves such as those where a turn arrow would be used when they should be signed more aggressively. Curves and Intersections Findings and recommendations regarding curves and inter- sections are as follows: • Notwithstanding the fact that all intersections were marked with Intersection Ahead signs, intersections in the curve caused difficulties for drivers. None of the signs “com- bined” the curve and intersection warning. It is not clear what more could or should be done, but it is clear that the combination is indeed problematic and should be signed whenever it occurs—that is, drivers should always be told when there is an intersection on or near a curve. Anecdotal Observations Based on DPM Subject Performance In addition to the more quantified results just recapitulated, a summary of more qualitative comments has also been pre- pared. The point of this exercise was to use the perspective of the DPM observer in assessing how drivers responded to curve-related TCDs and to the physical situations that they encountered in the field. 15 The drivers were observed to operate their vehicles “mechan- ically” in an incremental rather than continuous process and did not respond to TCDs as if they were messages about what was on the road ahead. The immediate responses were often “tokens” of response—for example, drivers would take their foot off the pedal although they would not actively slow down, in spite of a speed advisory being present. While in some instances this response was “good enough,” in others it was not. This is borne out by the quantifiable responses at sev- eral of the curves. If the curve was reasonably gentle and did not violate basic expectations, the drivers were in good shape. If the curve was a more serious undertaking, token responses then led to driver errors. The types of curves or situations where more drivers had errors included the following: • Curves where the drivers had limited or no visibility of the curves when the TCDs were first visible; • Curves where there were vertical curves, primarily hill crests that obscured the curve; and • When the curve was combined with another element, especially intersections. In general, although the curves were well-marked, the nominal responses to the TCDs (such as easing up on the gas pedal) were simply not enough when the situations were more complex. Consequently, a more positive or assertive response was delayed until the drivers actually saw the extent of the curve or other feature. In some instances, modest responses caused later driver errors. There was also a general “dilution” of the driver’s attention—that is, in addition to tending to the driving task, they also looked at the scenery, asked general questions, talked on a cell-phone, or asked spe- cific questions. This performance can be compared with the three trained MSP officers—they were highly focused on the task at hand and not easily distracted.