Information As A Source of Distraction

Chapter 7. The effect of frequency and spacing of guide signs on driver behavior

This study examined the effects of the frequency, spacing, and information content of guide signs on driver performance. Specifically, it looked at the effects of the number and spacing of supplemental guide signs and specific-service logo guide signs.

Highway agencies (e.g., State departments of transportation and toll authorities) are sometimes under pressure to post more supplemental guide signs than is permitted in the MUTCD.⁽²⁾ For instance, when a college or university is listed on a supplemental sign, other higher education institutions near the same interchange may request similar acknowledgment. The demand for listing on specific-service signs (i.e., signs for gas, food, lodging, camping, attractions, and 24‑h pharmacies) may also exceed available sign locations that conform to MUTCD requirements.

Supplemental guide signs provide information regarding destinations accessible from an interchange other than places displayed on the standard interchange signing. Like standard interchange guide signs, supplemental guide signs have white lettering on a green background. The MUTCD limits supplemental guide signs to one per interchange and a supplemental sign may list no more than two destinations.⁽²⁾ The manual also specifies that there should be at least 800 ft (244 m) between a supplemental sign and other guide signs but cautions that the supplemental sign may overload drivers’ ability to process information.

Specific-service guide signs have white lettering on a blue background. The MUTCD recommends that these signs be spaced a minimum of 800 ft (244 m) apart from each other and from other guide signs.⁽²⁾ The services on a specific-service sign may be represented by either text or a logo, with a maximum of six services per sign. A maximum of four specific-service signs may serve an interchange. The limitation on the number of guide signs, the distance between them, and the amount information on individual signs is intended to avoid overloading drivers’ ability to receive information and to make appropriate decisions.⁽²⁾

Although it seems reasonable that too much signing could decrease the effectiveness of navigation-related signing, there is little empirical evidence that this is the case, and no empirically based literature is available that would enable quantification of what is too much. NCHRP Report 488: Additional Investigations on Driver Information Overload presents a model for quantifying the relative information load of sign arrays along a highway segment but does not provide a method for determining the absolute level at which too much information is present.⁽²⁰⁾

NCHRP 488 describes driver information overload and its potential consequences as follows:⁽²⁰⁾

• “Driver information overload” is defined as providing a motorist with too much information, through a series of devices or conditions, for a driver to have adequate time to perceive and respond properly. Therefore, the information load on a driver is a property not only of the specific sign he or she is encountering, but also of the roadway context in which the sign occurs, the information context in which the sign occurs, characteristics of the driver, and the particular navigational task. Where drivers are confronted with more information than they can process, they may decelerate severely or drive unduly slowly, make late or erratic maneuvers, take an improper route alternative, ignore critical information, fail to monitor other traffic, or have excessive eyes-off-the-road time episodes. These behaviors have obvious safety and operational consequences. (p. 1)

A key point is that the information load from guide signs is not a property of the signs themselves but rather depends on the driver’s navigation task and the context in which the sign is encountered. Given that the driver requires navigation information, overload may occur if the driver cannot readily obtain, in the time available, the needed information on a sign or cannot determine that the needed information is not on that sign. The time available depends on the demands of traffic, roadway geometry, and the number of other signs that the driver needs to monitor. Drivers on a familiar route would not need navigation sign information and should not experience information overload attributable to guide signs. As long as other signs near the roadway, such as billboards, do not resemble highway signs, guide sign information overload should not occur.⁽⁵⁹⁾ In addition, because guide signs are color coded (i.e., blue background for specific-service signs and green background for other interchange and supplemental guide signs), drivers who need only one type of information (i.e., either services or destination guidance) should not be expected to be overloaded by the presence of both types of signing. Furthermore, legends on specific-service signs (e.g., lodging, gas) are intended to relieve the driver of searching all specific-service signs for a specific type of service. The use of color coding and service legends assumes drivers use this information to reduce the load imposed on them. With respect to specific-service sign legends, this experiment provided a test of that assumption.

Drivers who need navigation information may have insufficient time to acquire the information in the context of too many signs insufficiently spaced if the legibility distance is insufficient, if it is unclear which signs to attend to, or if the roadway and traffic demands are high.

Given the potential consequences of information overload, quantification of what is too much could be useful in guiding and supporting rule-making and guiding agencies that perceive a need to post guidance information that goes beyond current sign frequency and spacing guidelines and regulations.

The origin of the FHWA requirement for an 800-ft (244-m) minimum spacing between freeway guide signs is unclear. The first interstate freeway signing and marking manual, published by American Association of State Highway Officials (now AASHTO) in 1958, includes the provision that “in no case shall guide signs…be spaced closer together than 800 feet.”⁽⁶⁰⁾ Although the origin of the 800-ft (244-m) minimum is unclear, there is supporting justification for a distance of about that magnitude. One assumption behind the minimum distance between signs is that the most vulnerable (inexperienced and elderly) drivers can process only one sign at a time and the minimum spacing ensures that vulnerable drivers will not be simultaneously confronted with two signs that must be read. Another is that each sign must be available for a minimum amount of time so that the driver can time-share sign reading with other driving demands. Mace, Hostetter, and Seguin conducted a series of laboratory and on-road experiments and concluded that any exposure time of more than 2 s is adequate for processing highway guide signs.⁽⁵⁹⁾ Their subjects were not vulnerable road users so a minimum greater than 2 s is probably reasonable. Assuming a travel speed of 65mi/h (104km/h), 2 s would represent 190 ft (58 m) of travel distance. With 10-inch- (25-cm-) high letters, the minimum height for the smallest lettering on freeway signs, the legibility distance would be 300 ft (91 m) with an assumed minimum visual acuity of 20/40. Having read a guide sign and determining that a lane change maneuver is required, a maneuvering distance must be accounted for. The 18-inch (46-cm) letter height for critical freeway guide sign information yields a minimum legibility distance of 540 ft (165 m), and a maximum legibility distance of 1,080 ft (376 m). For a driver with 20/20 vision signs would need to be about 1,000 ft (305 m) apart to guarantee two guide signs would not be legible at the same time. Whether the presence of two legible guide signs less than 8.4 s apart (800 ft (244 m) divided by 95 ft/s (29m/s) confronts drivers with a time-sharing challenge (i.e., information overload) is an empirical question that has not been adequately addressed.

The purpose of this study was to explore the effect of varying numbers of guide signs and their spacing on driver performance. The effects of the following variables were addressed:

• Sign spacing of drivers’ performance in obtaining and using guide sign information.
• Number of destinations listed on guide signs on driver performance.
• Sign spacing on driver performance.
• Guide sign frequency on visual scanning of the road ahead.

The study examined driver performance in the case where the driver had information needs that relied on guide and specific-service signs.

Methods

Participants in this study drove on a simulated freeway with four lanes in their direction of travel. They were instructed to take freeway exits for Holt Avenue, Harvard University, and the Holiday Inn®. This instruction was intended to require participants to monitor all three types of signing: interchange guide signs, supplemental guide signs, and specific service signs. The freeway had 21 entrances and 22 exits that comprised 21 unique freeway segments, where a segment extended from exit gore to exit gore. Participants were not informed of how many exits were signed for the assigned destinations. There was one of each. The number of specific service signs per segment varied between zero and three. The number of specific service signs also varied from zero to three. There were always two advance guide signs per segment, but some exits had two street name destinations per sign while the remainder listed only one street name. The number of street names on the exit for Holt Avenue was a between-group variable with two levels (one or two street names). The number of supplemental guide signs per exit varied between one and three and was a with-group variable. The number of destinations per sign varied from one to two. The number of supplemental guide signs and the number of destinations at the exit for Harvard varied among subjects.

Dependent measures were the probability of taking correct and incorrect exits, the frequency and duration of various eye-glance measures, and speed and lane keeping in the proximity of guidesigns.

The Simulator

The experiment was conducted in the FHWA highway driving simulator. In the simulator experiments described earlier in this report, the simulator’s out-of-vehicle display consisted of a horizontal projection of a 240-degree field of view onto cylindrical screen. However, at the time of this study, the projectors had exceeded their life expectancy and were no longer capable of achieving the necessary brightness or resolution to support a sign study. Because the projectors could not be replaced within the timeframe of this study, three high-resolution LCD monitors were used to display the forward 104 degrees of the field of view. Figure 55 shows the view of the roadway on the LCD monitors. (The image on the monitors is from a subsequent study.) Twoof the original projectors were used to complete the side portions of the 240-degree horizontal display. Each of the LCDs was 30 inches (0.76 m) on the diagonal with a 16:10 aspect ratio. The LCD monitors’ resolution was 2,560 horizontal pixels by 1,600 vertical pixels. LCD brightness was approximately 108 fl (370 cd/m²⁾ with a typical contrast ratio of 1,000:1. The distance of the monitors from the driver’s eye point varied with driver height and seat position. The nominal distance of the center monitor was 36 inches (0.9 m). The right and left monitors were 39 and 49 inches (1 and 1.2 m) from the design eye point, respectively. All distance measurements were to the center of the respective displays. Images on each display were scaled to present a 1:1 correspondence with the real-world equivalents of the virtual world. All displays refreshed at 60Hz. The minimum pixel response time on the LCD was 8 ms.

Figure 55. Photo. View of the three LCD monitors from slightly behind the driver’s eye point.

Because the LCDs were relatively close to the driver’s eye point, the simulated horizon height was adjusted to account for eye-point height differences of individual participants.

The LCD monitors were mounted in the windshield area of the sedan in the simulator. This placement required removal of the windshield. The simulator’s motion base was not enabled in this experiment. The car’s instrument panel, steering, brake, and accelerator pedal all functioned in a manner similar to real-world compact cars. Rear view mirrors were simulated using 7.8‑inch-wide (20 cm) by 4.8-inch-high (12 cm) color LCD with 800 pixels horizontally and 480pixels vertically. These displays had a contrast ratio of 400:1. Left and right outside simulated mirrors were mounted over the sedan’s original outside mirrors. The center-mounted rear view LCD was placed as near as possible to the location of the vehicle’s original mirror.

The simulated vehicle was equipped with a hidden intercom system that enabled communications between the participant and a researcher who ran the experiment from a remote control room. The researcher in the control room could also view the face video from the eye-tracking system and thereby monitor the participant’s wellbeing.

The Simulation Scenario

Seven signing conditions were simulated across 21 freeway segments, where each segment had one entrance merge and one exit. Table 28 summarizes the key attributes of each condition. All conditions had three interchange guide signs; a 1-mi (1.6-km) advance sign, a 0.5-mi (0.8-km) advance sign, and a sign at the exit gore. Specific-service signs always preceded the 1-mi (1.6‑km) advance guide sign. The distances between each specific-service sign and the third specific-service sign with the 1-mi (1.6-km) advance sign are shown in the column labeled “Distance Between Signs” in table 28. With one exception, the supplemental guide signs were placed between the 1-mi (1.6-km) advance sign and the 0.5-mi (0.8-km) advance sign. In condition3, one supplemental sign followed the 1-mi (1.6 km) advance sign by 800 ft (244 m), and one supplemental sign followed the 0.5-mi advance sign by 800 ft (244 m). The distances between the 1-mi (1.6 km) advance sign and the other supplemental guide signs is specified in the “Distance Between Signs” column of table 28. The number of destinations applies to both the advance guide signs and the supplemental guide signs. Each destination appeared once across the 21segments.

Table 28. Summary of the seven signing conditions.

The Holt Avenue exit was always on one of the first seven segments, the Harvard University exit was always on one of the second series of seven segments (i.e., 8 through 14), and the Holiday Inn® exit was always on one of the final seven exits. There were seven different orders in which the seven signing conditions were presented, and the seven target exits occurred equally often in each condition. With the exception of 2 participants, who were inadvertently run in the wrong conditions, each of the 49 combinations of targeted exits with order of conditions was presented to 2 participants.

Guide sign lettering used the FHWA series E (modified) font with nominal 18-inch (46 cm) letter height. To achieve the appropriate legibility distance (800 to 1,000 ft (244 to 305 m) for the signs, all signs were oversized by a factor of 1.5. This resulted in letters subtending a visual angle equivalent to a 27-inch- (68-cm-) high lettering. The street names on the advance guide signs were all interchange street names from I-10 between Los Angeles, CA, and Ontario, CA. The destinations of the supplemental guide signs consisted of colleges, museums, and other destinations with no correspondence to actual geography (e.g., Cal Tech and the University of Virginia appeared on the same supplemental sign).

All specific-service signs displayed six logos. The logos on the specific-service signs, which were also oversized, were business logos used on signs on I-95 in Virginia. More than 30 unique logos were used. The MUTCD specifies that where lodging, food, and gas services are displayed, the signs are sequenced in that order.⁽²⁾ However, in this experiment when more than one service sign was present, the lodging services sign was always preceded by a food services sign. The reversal in the normal order was intended to make the search task more challenging. It was assumed that most motorists are unaware of the order specified by the MUTCD for service signs. The business logos on the signs varied from exit to exit as did the position of individual logos that appeared on multiple signs. The target logo (i.e., Holiday Inn®) was always in the center of the lower row on the lodging sign. Nominally, condition 1 did not include specific-service or supplemental guide signs. However, as was done for all participants, the research design called for condition 1 participants to take an exit signed for the Holiday Inn® and Harvard University. Therefore, for condition 1 participants, a lodging or supplemental guide sign was added at the intended exits so that all participants in all conditions would correctly take three exits if they made no navigation errors.

To make the driving workload somewhat realistic, a traffic simulation model was used to generate traffic. Although the same traffic was generated for all participants (the traffic model random number seed was constant), the location of individual participants in the traffic stream could and did vary depending on how fast the participant drove and in which lanes he or she traveled. Traffic counters were located between each exit, and the number of vehicles crossing a counter while the participant was between the respective exits was recorded. Table 29 shows the resulting statistics for vehicles per lane per hour during the time a participant was somewhere in the respective zones. Software failures resulted in some data loss as indicated by the Number of Participants column, which shows the number of participants contributing to each zone statistic.

Table 29. Vehicles/lane/h summary statistics for each data collection zone.

The simulated traffic did not enter or exit the freeway at any point. However, the simulated vehicles did change lanes and execute passing maneuvers. Participants who drove faster than 65mi/h (105 km/h) and attempted to closely follow other vehicles encountered more traffic than participants who maintained 65 mi/h (105 km/h) or less.

Table 30 shows the percent of vehicles in each lane for which each of four desired speeds was specified. The minimum desired speed in the right lane was set at 70 mi/h (113 km/h) so that participants who were instructed to maintain 65 mi/h (105 km/h) would not have an incentive to leave the right lane.

Table 30. Speed parameters (percent vehicles per speed bin per lane) for traffic control by the traffic simulation model.

Eye-Tracking System

The same eye-tracking system was used in this experiment as in the previous simulator experiments in this report. The mean (M) accuracy of left-eye gaze-position across participants was 1.3 degrees (radius) with an SD of 0.65 degrees (95th percentile accuracy M = 2.3, SD = 1.2). The mean accuracy of right-eye gaze-position across participants was 1.6degrees (radius) with an SD of 0.83 degrees (95th percentile accuracy M = 2.9, SD = 1.9). The eye-tracking data (e.g., gaze direction of each eye, head position, etc.) were merged with data from the simulator (e.g., vehicle speed, lane position, and steering wheel position) and the current forward view of the simulation visual scene (approximately 39 degrees horizontal by 24degrees vertical). The merge was accomplished using a MAPPS® scene recorder.⁽⁶¹⁾

To quantify when and for how long participants looked at each guide sign, a researcher used analysis software to indicate an ROI on individual frames of the recorded video image. An example of an analyst’s screen with a coded ROI for specific-service signs 400 ft (122 m) apart is shown in figure 56. Also shown is the road ahead ROI. When three specific-service signs were 200 ft (61 m) apart, their ROI overlapped with ROIs for the following advance guide sign and supplemental guide sign. Where ROIs overlapped, a set of priorities determined which ROI was credited to a gaze vector that landed on the overlap. The order of priorities, from highest to lowest, was specific service, supplemental, advance, and road ahead. These priorities had the greatest effect in condition 7, where the signs were 200 ft (61m) apart. Only conditions 6 and 7 had ROIs that included more than one sign, and only signs that were 200 ft (61 m) or 400 ft (122m) apart.

Each ROI included the sign and approximately 8 ft (2.4 m) (virtual) to each side of the sign. Each ROI was coded for 10 s on the approach to the sign and terminated when the sign began to pass out of the forward 39-degree field of view (i.e., was more than 19.5 degrees to the right driver’s forward view). For ROIs that included more than three signs, the coding began 10 s before reaching the first of the signs and ended when the last sign passed out of the field of view. Figure 57 shows an ROI for a specific-service sign, and figure 58 shows an ROI for an advance guide sign.

Figure 56. Screen capture. Analyst’s screen for coding shows ROI for specific-service signs that are 400 ft (122 m) apart.

Figure 57. Screen capture. ROI drawn around a specific-service sign.

Figure 58. Screen capture. ROI drawn around an advance guide sign.

Participants

Participants were recruited from a list of volunteers maintained by FHWA’s Safety Research and Development Human Factors Team. Of 127 individuals recruited, 113 completed driving the entire scenario. Of the 14 individuals who did not complete the drive, 2 dropped out because of simulator sickness, 2 elderly people had trouble with instructions or controlling the vehicle, 8persons either could not be calibrated for eye tracking or otherwise did not track well, and with 2 persons, the simulator hardware or software failed.

Of the 113 participants who completed the drive, 98 provided usable eye-tracking data. During recruitment, an attempt was made to balance the sample with respect to age and gender. Because the median age in the participant database was 46 years, the goal was to recruit equal numbers of male and female drivers over 46 and 46 years of age or younger. The mean age of the 98 participants with complete data was 44 years (range 18 to 76 years), with 49 females and 49males.

Instructions

The instructions to participants were as follows:

• You will be driving on a freeway that has four lanes in the direction you are traveling. There will be other vehicles on the road, although traffic should not be heavy. You should obey the law and drive as you normally would. You can drive in any lane you choose. The posted speed limit is 65 miles per hour. Your task is to watch for three destinations: Holt Ave, Harvard University, and the Holiday Inn. When you see the exit for any of these destinations, take that exit. We will be recording how you drive and where you look. Please drive as you normally would. If you travel close to 65 MPH and take all the correct exits, you will receive a $10 bonus. Do you have any questions? The trip should take about 45min.

  Remember, you are looking for:

  • Holt Ave
  • Harvard University
  • The Holiday Inn

Participants were also handed a picture of the Holiday Inn® logo so it could be assumed that the target logo was familiar.

Before beginning the test drive, participants completed a practice drive. The practice drive was similar to the test drive except that there was no traffic and there were only three exits. The exits on the practice drive were signed for the Holiday Inn®, Harvard University, and Holt Avenue and were presented in that order. Each of the practice exits had three guide signs, one specific-service sign, and one supplemental guide sign with spacing compliant with the MUTCD guideline.⁽²⁾

Participants received $40/h compensation for a typical 1.5 h of participation. In addition, the
$10 bonus mentioned in the instructions was paid when appropriate.

Results

Throughout this report, error bars in the charts and graphs represent 95-percent confidence limits around the means.

Exit Taking

Table 31 shows the percent of participant results for the first seven exits, one of which contained advance guide signs with the Holt Avenue destination. Recall that each condition was represented once among the first seven segments. Therefore, for each condition, a correct response is indicated on the positive diagonal (i.e., no/no and yes/yes cells), if guide signs did not include Holt Avenue, then the correct response was not to exit (no/no), and if the guide signs listed Holt Avenue, then the correct response was to exit (yes/yes). If the participant took an exit that did not have Holt Avenue as a destination, then that participant is represented in the upper-right cell for that condition (no/yes), and if the participant failed to take the Holt Avenue exit, then that participant is represented in the lower-left cell for that condition.

Table 31. Percent of participants on segments 1–7 who responded correctly to guide signs (no/no and yes/yes) and false alarms (no/yes) and misses (yes/no) as a function of signing condition.

Table 32 shows the analogous results for segments 8–14, where the correct exit had a supplemental guide sign that listed Harvard University as a destination. Table 33 summarizes exit-taking performance for segments 15–21, where the correct exit had a supplemental guide sign with the Holiday Inn® logo.

Table 32. Percent of participants on segments 8–14 who responded correctly to the supplemental guide signs (no/no and yes/yes) and false alarms (no/yes) and misses (yes/no) as a function of signing condition.

Table 33. Percent of participants on segments 15–21 who responded correctly to the specific-service signs (no/no and yes/yes) and false alarms (no/yes) and misses (yes/no) as a function of signing condition.

There were very few exit-taking errors, and the errors that were observed did not appear to be associated with a type of sign (i.e., street name, supplemental destination, and specific service) or with any of the variables associated with the seven signing conditions (i.e., sign spacing, number of signs, or number of destinations on the signs). It is important to remember that the participants were not informed which exits might have which target destination and therefore were expected to search for all 3 targets at all 21 exits. The distinction between the three sets of seven exits is made so that if there were a difference in the detection rate of a particular type of target, that could be seen. Also it is evident that the target detection rate was stable across the 21 exits.

The research design was not factorial with respect to the number of signs, the number of destinations on a sign, or the distance between signs. Therefore, pairwise comparisons were used to test for significant differences with respect to individual factors and interactions between factors. For instance, to test whether the distance between signs affected correct exit taking, McNemar’s Test was performed on the data in table 34. The No/No cell is empty because no participant made an exit error in both condition 6 (400 ft (122 m) between signs) and condition 7 (200 ft (61 m) between signs). One participant (1.02 percent) made no errors in condition 7 and at least one error in condition 6. Six participants made at least one error in condition 7 but none in condition 6. Ninety-one participants made no errors in either condition 6 or 7. Although there were more errors with 200-ft (61-m) sign separation than with 400-ft (122-m) separation, this difference is small and not statistically significant at the 0.05 level.

Table 34. Percent of participants who responded in either conditions 6 or 7 or both 6 and 7.

In all cases, McNemar’s Test assuming a binomial distribution was used.⁽⁶²⁾ No significant differences in correct exit-taking behavior were observed as a function any of the three factors or their interactions.

Gaze Behavior

Only glance and fixation measures were used in the analysis of gaze behavior in this study. Looks were not addressed for two reasons: (1) in the previous experiment, the look measure led to conclusions similar to those for the glance and fixation measures, and (2) where gaze vectors fell near the border between ROIs, it was difficult to determine where one look ended and another began because the vector may frequently shift back and forth across the border between ROIs.

Glance Results

As in the previous simulator experiments, a glance was defined as any accumulation of 12 (12 × 0.008 s = 100 ms) or more hits on an ROI, where a hit is a single 120-Hz center of gaze vector estimate that falls on the sign’s ROI.

Table 35 shows the probability of a glance to sign ROIs. Where there was more than one ROI for a sign type (e.g., first and second advance guide signs and the (third) exit guide sign), the probabilities for each ROI are shown. Cells are marked “N/A” if there was no second or third ROI. In conditions 5 and 6, there were three specific-service signs and three supplemental guide signs, but only one ROI that cover all three signs of the respective types. The probability of glancing at all unique signs was high and never less than 0.86. The probability of glancing at redundant second and third guide signs tended to be lower than for other signs but was never less than 0.74.

Table 35. Probability of glance at signs as a function of condition and sign type.

Although the two advance guide signs and the guide sign at the exit contained redundant information, drivers tended to glance at each sign regardless of glances at the previous sign. Table 36 shows the conditional probability of glances at a second advance guide sign given that the first sign received a glance and the probability of a glance at an exit guide sign given a glance at the preceding (0.5 mi (0.8 km) advance guide sign.

Table 36. Probability of glancing at a subsequent guide sign given a glance at the preceding advance guide sign.

The mean duration of glances to sign ROIs was calculated two ways. First, the average duration of glances was computed for all ROI instances where zero duration (actually 0.00001 s) was recorded when drivers did not glance at an ROI. This method provided an estimate of the total attention a sign might attract and is shown in table 37. Cells in table 37 marked “N/A” represent locations where signs did not exist. Second, the average durations of glances at an ROI was computed only for actual glances (i.e., mean duration of a glance given a glance took place). These means are shown in table 38. Recall that glance durations are the sum of all 0.0083-s hits of a gaze vector on an ROI and may include multiple fixations interspersed with glances elsewhere. Specific service-sign ROIs captured considerably longer glances, more than 3 s in all conditions except for the gas sign in condition 5, which still had an average glance duration of more than 2 s. In condition 6, three specific-service signs, 400 ft (122 m) apart, were included in one ROI; that ROI had a mean glance duration of 7.4 s. Assuming a speed of 95 ft/s (20 m/s), the ROI was present for 18 s. Thus for that 18 s, the sign captured visual attention 41 percent of the time, on average. The supplemental guide signs in condition 6 also received a high proportion of visual attention, about 34 percent of the 18 s exposure to that ROI. The specific-service and supplemental guide signs in condition 7 were 200 ft (61 m) apart, which resulted in an approximate exposure duration of 14 s rather than the 18 s in condition 6. For the 14 s that glances at the specific-service sign ROI could be recorded, the signs captured visual attention 42 percent of the time on average.

Table 37. Mean glance duration in seconds where 0 s was used for no glance.

Table 38 provides mean glance durations without averaging in zero durations. As would be expected, these means were slightly longer than those shown in table 37. However, the patterns remained the same. Specific-service signs garnered glances more than twice as long as those directed at guide signs, and guide sign glance durations were brief. Where there were two destinations on advance and supplemental guide signs (conditions 3 and 5), glance durations were not substantially greater than when there was only one destination (conditions 1, 2, and 4).

Table 38. Mean glance duration in seconds for sign ROIs given a glance occurred.

To further explore glance behavior, each 0.0083-s gaze vector was assigned to one of threeobject categories: signs, road ahead, and other. The other category included all gaze vectors with the forward 39- by 24-degree scene that were not on a sign or road ahead ROI. Such hits could be toward other driving-related information (e.g., traffic) or non-driving-related information (e.g., billboards). Hits on any guide sign, supplemental guide sign, or specific-service sign were included in the sign category.

The proportion of gaze to the three gaze categories was analyzed twice. The first analysis included the entire distance between the on-ramp merge area and exit gore. The second analysis focused on the area around the first advance guide sign so that any effect that resulted from increasing the number of signs and decreasing the space between signs would be maximized. Specifically, this second analysis spanned an area that began at the end of the entrance merge area and ended with last glance at a sign ROI before reaching the second advance guide sign.

The proportions of gaze were analyzed using a GEE model that assumed a gamma response distribution and identify link function. The response variable was proportion of time. Predictors in the model included object category (sign, road ahead, or other), signing condition, and the object-condition interaction. Significance was evaluated with Wald χ² statistics. Because the research design was not factorial, conditions were analyzed in pairwise fashion. The number of pairwise comparisons was large, which could increase the probability of type 1 errors
(i.e., finding a significant effect due to chance variation in means that are not different in the population). To protect against this increase in type 1 errors due to alpha inflation, a p criterion of 0.002 was used rather than the traditional 0.05.

Condition 1 Versus 2—Effect of Addition of Specific-Service and Supplemental Guide Signs:

Condition 1 had no supplemental guide signs or specific-service signs except 2 participants for whom 2 out of 21 exits were designated with one of these signs. Condition 2 had one lodging sign and one supplemental guide sign, and the guide signs listed just one destination. Therefore, condition 1 represent a test of the effect of adding a specific-service and supplemental guide sign.

Category ( χ²(2) = 201.00, p < 0.001), condition ( χ²(1) = 96.17, p < 0.001), and the interaction ( χ²(2) = 105.38, p < 0.001) were statistically significant. Because the interaction was significant, main effects are not discussed. Table 39 displays summary statistics for the comparison of conditions 1 and 2 where the response variable covered the entire data collection zone.

Table 39. Comparison of the GEE estimates proportion of gaze time between conditions 1 and 2 for the entire data collection zone.

As can be seen in table 40, the proportion of time to other was relatively constant, whereas there was a tradeoff between the proportion of time to road ahead and signs: proportion of time to road ahead decreased from condition 1 to 2, and the proportion of time to signs increased.

As with the full data collection zone analysis, when only the area around the first advance sign is considered, category ( χ²(2) = 150.80, p < 0.001), condition ( χ²(1) = 82.06, p < 0.001), and the second-order interaction ( χ²(2) = 129.10, p < 0.001) were significant. In this second analysis, the same tradeoff between road ahead and signs is evident, but more pronounced.

Table 40. Comparison of the GEE estimates of the proportion of gaze between conditions 1 and 2 for the area around the first advance guide sign.

Condition 2 Versus 3—Effect of Adding a Second Destination to Guide Signs:

Conditions 2 and 3 had the same number of signs, but the guide and supplemental guide signs in condition 3 listed two destinations. As can be seen in table 41 and table 42, the effect of adding destinations to the guide signs was to decrease the proportion of gaze time to the road ahead and perhaps to other. The interaction between category and condition was significant for both the entire data collection zone, χ²(2) = 51.54, p < 0.001, and the portion of the data collection zone around the first guide sign, χ²(2) = 34.74, p < 0.001.

Table 41. Comparison of the GEE estimates proportion of gaze between conditions 2 and 3 for the entire data collection zone.

Table 42. Comparison of the GEE estimates of the proportion of gaze between conditions 2 and 3 for the area around the first advance guide sign.

Condition 2 Versus 4—Effect of Adding a Second Supplemental Guide Sign:

The difference between conditions 2 and 4 was that condition 2 had only one supplemental guide sign, whereas condition 4 had two supplemental guide signs. All signs were 800 ft (244 m) apart with both supplemental guide signs coming after the first advance guide sign.

As can be seen in table 43, there was a slight increase in the proportion of gaze time at the signs and a slight decrease in the proportion of time at the road ahead when there were two supplemental guide signs rather than one. This resulted in a significant interaction of ROI category and condition, χ²(2) = 17.33, p < 0.001. When only the partial data collection zone was considered, the interaction effect was not statistically significant.

Table 43. Comparison of GEE estimates of the proportion of gaze between conditions 2 and 4 for the entire data collection zone.

Condition 3 Versus 5—Addition of a Second Specific-Service Sign and a Second Supplemental Guide Sign:

Conditions 3 and 5 both had two destinations listed on the guide signs. The difference between conditions 3 and 5 was that condition 3 had one specific-service sign and one supplemental guide sign, whereas condition 5 had two signs of each of those types. The second supplemental guide sign came 800 ft (244 m) after the second advance guide sign.

As can be seen in table 44, there was a tradeoff between the proportion of gaze time at the road and at signs. With the increase in the number of specific-service and supplemental guide signs, the signs captured more visual attention at the expense, primarily, of gaze at the road ahead. The interaction between condition and category was significant, χ²(2) = 151.85, p<0.001.

Table 44. Comparison of GEE estimates of the proportion of gaze between conditions 3 and 5 for the entire data collection zone.

The partial data collection zone did not include the second supplemental guide sign. Thus, the partial data collection zone comparison was actually a test of the effect of adding a second specific-service sign.

As can be seen in table 45, signs captured about 6-percent more of the proportion of gaze when there were two specific-service signs rather than one. This increase in the proportion of gaze at the signs came at the expense of both the other and road ahead categories. The interaction between category and condition was statistically significant, χ²(2) = 117.79, p < 0.001.

Table 45. Comparison of GEE estimates of the proportion of gaze between conditions 3 and 5 for the partial data collection zone.

Condition 5 Versus 6—Effect of Adding a Third Specific Service and Supplemental Guide Signs and Reducing Distance Between These Signs to 400 ft (122 m):

Because the data collection zones were all the same length, adding a third specific-service sign and a third supplemental guide sign required reducing the distance between signs. In condition 4, the signs were all 800 ft (244 m) apart and thus were compliant with the MUTCD sign spacing requirement.⁽²⁾ In condition 5, all signs prior to the second advance service sign were 400 ft (244m) apart. Two destinations were listed on all guide signs in both conditions. The third specific-service sign was a gas sign that followed the lodging sign.

As can be seen in table 46, increasing the number of signs and decreasing the distance between signs increased the proportion of gaze at the signs at the expense of both the road ahead and other. The interaction between category and condition was statistically significant, χ²(2) = 133.73, p < 0.001. All of the differences between conditions 5 and 6 were in the partial data collection zone around the first advance guide sign. As can be seen in table 47, the proportion of gaze at the signs in this area increased by about 14 percent at the expense of both the road ahead (5 percent) and other (10 percent). The category by condition interaction was statistically significant, χ²(2) = 280.81, p < 0.001.

Table 46. Comparison of GEE estimates of the proportion of gaze between conditions 5 and 6 for the entire data collection zone.

Table 47. Comparison of the GEE estimates of the proportion of gaze between conditions 5 and 6 for the partial data collection zone.

Condition 6 Versus 7—Effect of Reducing Distance Between Signs from 400 ft (122 m) to 200 ft (61 m):

Conditions 6 and 7 were the same except for the distance between signs. In condition 7, the signs were 200 ft (61 m) apart. Because the differences between conditions were in the partial data collection zone, only the partial data collection zone results are presented. As can be seen in table 48, decreasing the distance between signs decreased the proportion of gaze at the signs. This effect may be because the data collection zone was only 600 ft (183 m) shorter in condition7 than in condition 5, but the area containing the signs was 1,200 ft (366 m) smaller in condition 7. Thus, in condition 7, the proportion of time the signs were legible was reduced compared with condition 6. The reduction in the proportion of gaze time at the signs from condition 6 to 7 was statistically significant as indicated by the significant interaction between category and condition, χ²(2) = 44.28, p < 0.001.

Table 48. Comparison of GEE estimates of the proportion of gaze between conditions 6 and 7 for the partial data collection zone.

Summary of Paired Gaze Comparisons:

As signs were added and the space between signs was reduced, the proportion of time attending to the road ahead, as defined by the road ahead ROI, went down, and the proportion of gaze to signs increased. Table 49 shows how the proportion of time that the gaze vector fell on signs as a function of condition. Consistent with the duration of gaze data in table 38, the biggest increases in gaze at signs came when specific-service logo signs were added: from none to one (conditions 1 to 2), from one to two (condition 2 to 5), and from two to three (conditions 5 to 6).

Table 49. Summary of findings for proportion of gaze to signs.

Fixation Results

As in the previous simulator experiments, a fixation was defined as seven consecutive gaze positions (60 ms) within a fixation radius of 4 percent of the vertical image height (i.e., 29 pixels on a 720-pixel image) and centered on the first of the seven gaze positions that designated the start of a fixation. The fixation continued until there were six consecutive hits (50 ms) outside the fixation radius. For a simulated object 500 ft (152 m) ahead, the fixation radius subtended a visual angle of about 2 degrees. A fixation on an ROI was recorded if the center of the fixation was on the ROI.

The mean probability of at least one fixation falling on a sign ROI is shown in table 50. The probability of fixating on any particular guide sign (advance or exit) was somewhat lower than for specific-service and supplemental signs, perhaps because participants realized the guide sign information was redundant across the three signs and fixations on any one or two of the guide signs could be sloughed off without loss of information.

Table 50. Probability of fixating on each ROI category as a function of condition.

The mean number of fixations as a function of condition and ROI category is shown in table 51. Note that the number of fixations on single lodging service signs is roughly twice that for supplemental guide signs, and food and gas signs received only slightly fewer fixations than the lodging signs. Table 52 shows the mean duration of each fixation on an ROI. For the specific-service signs in condition 6, the product of the average number of fixations multiplied by the average duration of fixations yields a total duration of fixation on that ROI of 7.68 s. With two specific-service signs in condition 4, the sum of the average fixation duration times their respective mean frequencies was 5.81 s. For a single specific-service sign in condition 2, the total fixation duration was 3.65 s. Although participants only needed to search the lodging sign, it is evident that food and gas signs added to their search time.

Table 51. Mean number of fixations as a function of condition and ROI category.

Table 52. Mean fixation duration on ROIs given at least one fixation.

The proportion of fixations on the three categories of ROI, sign, road ahead, and other were analyzed in the same manner as the gaze vector hits on ROIs. As before, a GEE model was used that assumed a gamma distribution and identify link function. Significance was evaluated with a 0.002 alpha criterion.

Condition 1 Versus 2—The Effect of Adding One Specific-Service and One Supplemental Guide Sign:

As can be seen in table 53, adding two signs—a lodging sign and a supplemental guide sign—more than doubled the number of fixations on signs and reduced fixations on the road ahead. This interaction effect was statistically significant, χ²(2) = 89.63, p < 0.001. When looking only at the partial data collection zone, the results were similar, with a higher proportion of fixations on the signs compared with the road ahead, and the interaction effect was again significant.

Table 53. Estimated mean proportion of fixations on signs, road ahead, and other as a function of signing condition over the entire data collection zone.

Condition 2 Versus 3—The Effect of Adding a Second Destination:

As can be seen in table 54, which shows the proportion of fixations over the entire data collection zone, the addition of a second destination on the guide signs added about 3 percent to the proportion of fixations on signs and reduced the proportion of road ahead fixations by a similar amount. This resulted in a significant category by condition interaction, χ²(2) = 35.95, p < 0.001.

Table 54. Estimated mean proportion of fixations on ROI categories as a function of signing conditions 2 and 3 over the entire data collection zone.

Condition 2 Versus 4—The Effect of Adding a Second Supplemental Guide Sign:

As can be seen in table 55, which shows the proportion of fixations over the entire data collection zone, the addition of a second supplemental guide sign added about 2 percent to the proportion of fixations on the signs at the expense of fixations on the road ahead and other. This resulted in a significant category by condition interaction, χ²(2) = 19.30, p < 0.001.

Table 55. Estimated mean proportion of fixations on ROI categories as a function of signing conditions 2 and 4.

Condition 3 Versus 5—The Effect of Adding a Second Specific-Service and Second Supplemental Guide Sign:

As can be seen in table 56, which shows the proportion of fixations over the entire data collection zone, the addition of a second specific-service sign and a second supplemental guide sign added about 6 percent to the proportion of fixations on the signs at the expense of fixations on the road ahead. This resulted in a significant category by condition interaction, χ²(2)= 168.56, p < 0.001.

Table 56. Estimated mean proportion of fixations on ROI categories as a function of signing conditions 3 and 5.

Condition 5 Versus 6—The Effect of Adding a Third Specific-Service Sign and a Third Supplemental Guide Sign and Reducing Distance Between Signs to 400 ft (122 m):

As can be seen in table 57, which shows the proportion of fixations over the entire data collection zone, the addition of a third specific-service sign and a third supplemental guide sign added about 7-percent more fixations to the signs at the expense of fixations on other and the road ahead. This resulted in a significant category by condition interaction, χ²(2) = 173.95, p < 0.001.

Table 57. Estimated mean proportion of fixations on ROI categories as a function of signing conditions 5 and 6.

Condition 6 Versus 7: The Effect of Reducing the Distance Between Signs From 400 ft (122m) to 200 ft (61 m):

As can be seen in table 58, which shows the proportion of fixations over the entire data collection zone, the reduction in the distance between signs resulted in a 4-percent decrease in the proportion of fixations on signs to the benefit of the road ahead and other. This resulted in a significant category by condition interaction, χ²(2) = 101.26, p < 0.001.

Table 58. Estimated mean proportion of fixations to the ROI categories as a function of signing conditions 6 and 7.

Although the trends and statistically significant effects were the same for the proportion of fixations in the full data collection zone and the partial data collection zone, the magnitude of the reduction in fixations on the road ahead is more striking when only the area around the first advance guide sign is considered. In the partial data collection zone for condition 6 (see table 59), which included more than 3,300 ft (396 m) of travel distance, the road ahead captured only 34 percent of fixations compared with the signs, which captured, on average, almost half the fixations.

Table 59. Partial data collection zone estimates of the mean proportion of fixations on ROI categories as a function of signing conditions 6 and 7.

Driving Performance Measures

Measures of the steering wheel position were not available for this study because of data collection problems related to a change in the driving simulator hardware.

Vehicle speed was analyzed for the partial data collection zones. It was hypothesized that if the signs caused information overload, drivers might slow to give themselves more time to search for the assigned destinations. On average, participants drove at the instructed speed of 65 mi/h (105 km/h). GEE models were used to test for difference in speed as a function of condition. A Gaussian response distribution with identity link function was assumed. Using an alpha level of 0.05, which does not adjust for experiment-wise error rate, there were two significant reductions in speed that were consistent with the hypothesis that information demand caused a speed reduction. The planned comparisons that were examined are shown in table 60. Estimated mean speed in condition 6 was significantly slower than estimated mean speed in condition 5, χ²(1) = 5.36, p=0.021, and estimated mean speed in condition 2 was significantly faster than estimated mean speed in condition 7, χ² (1) = 4.89, p = 0.027.

Table 60. Planned between-condition vehicle speed comparisons.

Table 61 shows the estimated mean speed in the partial data collection zones as a function of signing condition. Although statistically significant differences were identified, all differences were less than 1 mi/h (1.61 km/h).

Table 61. Vehicle estimated mean speed (in mi/h) in the partial data collection zones as a function of signing condition.

Discussion

Participants in this experiment were given a task that is probably rare in the real world—to watch for a guide sign destination, a supplementary guide sign destination, or a hotel on a specific-service sign. Both the eye-tracking data and the exit-taking behavior suggest that the participants took the task seriously. The mean speed data suggest, however, that while monitoring the signs for the assigned destination, the participants continued to monitor the driving task sufficient to maintain the instructed speed.

The eye-tracking results appear to support the current MUTCD standards and guidance on the frequency and spacing of guide signs.⁽²⁾ All additional signing resulted in increased eyes-off-road time. In particular, specific-service signs appear to be problematic. All specific-service signs had six logos. On average, participants fixated on a lone lodging sign for 3.65 s, with an average of about eight fixations that averaged 0.4 s each. Each additional logo sign added about 2 s of total fixation time, even though the additional signs were not lodging signs and did not require a search beyond noting the panel legend. When the fixation data were scored, it appeared that most participants searched all the specific-service signs for the Holiday Inn® logo. This suggests that the sign legends (i.e., food, lodging, and gas) had a minimal effect on search strategy.

The difficulty participants had in searching the logo signs may also be related to the distinguishability of the logos. Recent research that addressed the question of the maximum number of logos on a single specific-service sign used a reaction time paradigm to determine the amount of time it takes to detect a desired business logo on signs with four, six, or nine logos.⁽⁶³⁾ Participants’ only task was to watch a video monitor and strike a key labeled “yes” if a particular logo was on the sign or a key labeled “no” if it was not. From the time the sign appeared until the participants pressed one of keys was the dependent measure. Mean reaction times for participants less than 50 years of age were 1.3 s, 1.6 s, and 2.2 s for four-, six-, and nine-panel signs, respectively. Based on the generally cited criterion that glances away from the road of more than 2 s are unsafe, the investigators recommend that nine-panel logo signs not be approved.^(64,20) Two key assumptions underlie this recommendation. One is that looking at logo signs constitutes a glance away from the roadway. For most of the 10-s approach to the logo signs for which fixation data were collected in the present study, the signs were within 10 degrees of the participants’ forward field of view. In naturalistic driving studies, these would not be considered glances away from the forward roadway, because in naturalistic driving studies, which to date have not used eye tracking technology, gaze direction measurement is accurate to a radius of 20degrees.⁽⁶⁵⁾ By this criterion, 3.65 s might not be an unsafe amount of time to gaze at a roadside sign. Furthermore, although total fixation and gaze times at the specific-service signs were long, individual glances away from the roadway were general short—less than 0.5 s.

The other key assumption of the laboratory reaction time study was that the reaction time for deciding whether a particular logo is on a sign is an indicator of the amount of time a driver on the roadway would search for a logo on a sign. This assumption may be true in a relative sense, (i.e., longer laboratory reaction times indicate longer on-road search time). The present results suggest that the laboratory reaction time findings are not indicative of the absolute search time in a driving context. Search times on the six-panel signs in this study were more than twice the decision reaction times in the laboratory.

Whether search times of 3.7 s (one six-panel sign), 5.8 s (two six-panel signs) or 7.7 s (three six‑panel signs) represent a distraction is not clear, although the search times should raise concerns, particularly when they represent close to half of the fixation time over more than 0.5mi (0.8 km) of travel distance. A simulator study that presents various roadway hazards while participants search for specified logos might clarify this issue.

It is possible that with more distinctive logos, search times could be reduced. The target logo in the present study was light green on a white background. There were several logos, although none were lodging logos, with similar white backgrounds and low-contrast green images. However, many businesses have logos with low contrast and fine details that are not easily identified. Some national chains have logos that may be quickly identified at distances of 1,000ft (305 m), but these are not the norm. In that sense, the present search task was a representative one. The MUTCD does not control the design of business logos, only the maximum size of logo panels on specific-service signs.⁽²⁾

Total fixation durations (number of fixations multiplied by mean fixation duration) for supplemental guide signs were less than 2 s when there was one supplemental sign containing either one or two destinations. With two supplemental guide signs with a single destination on each, the sum of fixation times was about 0.5 s longer than for two destinations on one sign. This suggests that combining two destinations on one sign, the recommended practice in MUTCD, is a better choice than using two signs for those destinations.⁽²⁾ When there were two or more supplemental guide signs with two destinations on each, total mean fixation time ranged between about 4 and 6.3 s.

Summary and Conclusions

The MUTCD standards for supplemental guide signs are strongly supported by this research.

The findings with respect to specific-service signs are more problematic. The results certainly support not exceeding the current standard of no more than six logos per sign and no more than four signs with minimum separations of 800 ft (244 m). Methods for increasing the conspicuity of specific-service sign legends should be explored so that drivers do not unnecessarily search signs that list services that they are not seeking.

References

1. Ullman, G.L., Ullman, B.R., Dudek, C.L., and Trout, N.D. (2004) Legibility Distances of Smaller Character Light-Emitting Diode (LED) Dynamic Message Signs for Arterial Roadways, Texas Transportation Institute, College Station, TX.
   
   
2. FHWA. Manual on Uniform Traffic Control Devices for Streets and Highways. [Available online at http://mutcd.fhwa.dot.gov/pdfs/2009/mutcd2009edition.pdf.] Accessed June 17, 2014.
   
   
3. Dudek, C.L. (2004) Changeable Message Sign Operation and Messaging Handbook, FHWA-OP-03-070. FHWA, Washington, DC.
   
   
4. Nygardhs, S. (2011) Literature Review on Variable Message Signs: VMS 2006–2009, VTI notat 15A-2011. VTI, Linkoping, Sweden.
   
   
5. Dudek, C.L. and Ullman, G.L. (2001) Variable Message Sign Operations Manual Final Report, FHWA-NJ-2001-10. New Jersey Department of Transportation, Trenton, NJ.
   
   
6. Regan, M.A., Lee, J.D., and Young, K.L. (2008) “Chapter 1. Introduction.” Driver Distraction: Theory, Effects, and Mitigation, M.A. Regan, J.D. Lee, and K.L. Young, Eds., CRC Press, Boca Raton, FL, p. 3–7.
   
   
7. US-EU Bilateral ITS Technical Task Force. (2010) Expert Focus Group on Driver Distraction: Definition and Research Needs. Berlin, Germany.
   
   
8. Horberry, T. and Edquist, J. (2008) “Distractions Outside the Vehicle.” Driver Distraction: Theory, Effects, and Mitigation, M.A. Regan, J.D. Lee, and K.L. Young, Eds., CRC Press, Boca Raton, FL, p. 215–227.
   
   
9. Young, K.L., Regan, M.A. and Lee, J.D. (2008) “Measuring the Effects of Driver Distraction.” Driver Distraction: Theory, Effects, and Mitigation, Regan, M.A., Lee, J.D., and Young, K.L., Eds., Boca Raton, FL, CRC Press, p. 85–105.
   
   
10. Corbetta, M., Akbudak, E., Conturo, T., Snyder, A., Ollinger, J., Drury, H., Linenweber, M., Petersen, S., Raichle, M., Van Essen, D., and Shulman, G.A. (1998) “Common Network of Functional Areas for Attention and Eye Movements.” Neuron, 21, p. 761–773.
    
    
11. Posner, M.I. (1980) “Orienting of Attention.” The Quarterly Journal of Experimental Psychology, 32, p. 3–25.
    
    
12. Victor, T.W., EngstrÖm, J., and Harbluk, J.L. (2008) “Distraction Assessment Methods Based on Visual Behavior and Event Detection.” Driver Distraction: Theory, Effects, and Mitigation, Regan, M.A., Lee, J.D., and Young, K.L., Eds., CRC Press, Boca Raton, FL, p.135–165.
    
    
13. Bhise, V.D. and Rockwell, T.H. (1973) Development of a Methodology for Evaluating Road Signs, OHIO-DOT-01-73. Ohio State University, Columbus, OH.
    
    
14. Sprenger, A., Schneider, W., and Derkum, H. (1999) “Traffic Signs, Visibility and Recognition.” Vision in Vehicles VII, Gale, A.G., Brown, I.D., Haslegrave, C.M., and Taylor, S.P., Eds., Elsevier, Amsterdam, The Netherlands, p. 421–425.
    
    
15. AhlstrÖm, C., Dukic, T., Ivarsson, E., Kercher, A., Rydbeck, B., and VistrÖm, M. (2010) Performance of a One-Camera and a Three-Camera System, Vip-publication 2010-1. Virtual Prototyping and Assessment by Simulation, Goteborg, Sweden.
    
    
16. Victor, T.W., Harbluk, J.L., and EngstrÖm, J.A. (2005) “Sensitivity of Eye-Movement Measures to In-Vehicle Task Difficulty,” Transportation Research Part F: Traffic Psychology and Behaviour, 8, p. 167–190.
    
    
17. Östlund, J., Nilsson, L., Carsten, O., Merat, N., Jamson, H., Jamson, S.L., Mouta, S., Carvalhais, J., Santos, J.A., Anttila, V., Sandberg, H., Luoma, J., De Waard, D., Brookhuis, K., Johansson, E., Engstrom, J., Victor, T.W., Harbluk, J.L., Janssen, W., and Brouwer, R. (2004) Haste Deliverable 2: HMI and Safety-Related Driver Performance, HASTE-D2. Institute for Transport Studies, Leeds University, Leeds, UK.
    
    
18. FHWA. 23 Code of Federal Regulations Part 655. [Available online at http://edocket.access.gpo.gov/2009/pdf/E9-28322.pdf.] Accessed July 10, 2015.
    
    
19. Kling, J.W. and Riggs, L.A., Eds. (1971) Experimental Psychology, Holt, Rinehart and Winston, New York, NY.
    
    
20. Lerner, N.D., Llaneras, R.E., McGee, H.W., Taori, S., and Alexander, G. (2003) Additional Investigations on Driver Information Overload, NCHRP Report 488. National Academy Press, Washington, DC.
    
    
21. Erke, A., Sagberg, F., and Hagman, R. (2007) “Effects of Route Guidance Variable Message Signs (VMS) on Driver Behaviour,” Transportation Research Part F: Traffic Psychology and Behaviour, 10, p. 447–457.
    
    
22. Jamson, S.L., Tate, F.N., and Jamson, A.H. (2005) “Evaluating the Effects of Bilingual Traffic Signs on Driver Performance and Safety,” Ergonomics, 48, p. 1,734–1,748.
    
    
23. National Electrical Manufacturers Association. (2005) Hardware Standards for Dynamic Message Signs (DMS) with NTCIP Requirements, National Electrical Manufacturers Association, Arlington, VA.
    
    
24. Joint Committee on the NTCIP. (2010) DMS Object Definitions, Version 03, AASHTO, ITE, NEMA, Washington, DC.
    
    
25. Garvey, P.M. and Mace, D.J. (1996) Changeable Message Sign Visibility, FHWA-RD-94-077. FHWA, Washington, DC.
    
    
26. Simlinger, P., Egger, S., and Galinsk, C. (2008) Proposal on Unified Pictograms, Keywords, Bilingual Verbal Messages and Typefaces for VMS in the TERN. IN-SAFETY-Infrastructure and Safety Project, International Institute for Information Design, Vienna, Austria.
    
    
27. Garvey, P.M., Pietrucha, M.T., and Meeker, D. (1997) “Effects of Font and Capitalization on Legibility of Guide Signs.” Transportation Research Record 1605, p. 73–79. Transportation Research Board, Washington, DC.
    
    
28. Holick, A.J., Chrysler, S.T., Park, E.S., and Carlson, P.J. (2006) Evaluation of the Clearview Font for Negative Contrast Traffic Signs, FHWA/TX-06/0-4984-1. Texas Department of Transportation, Austin, TX.
    
    
29. McDougall, S., Tyrer, V., and Folkard, S. (2006) “Searching for Signs, Symbols, and Icons: Effects of Time of Day, Visual Complexity, and Grouping.” Journal of Experimental Psychology: Applied, 12, p. 118–128.
    
    
30. Wang, J.-H., Collyer, C.E., and Clark, A.T. (2007) Assisting Elder Drivers' Comprehension of Dynamic Message Sign Messages, URITC FY 06. University of Rhode Island Transportation Center, Kingston, RI.
    
    
31. Dudek, C.L., Schrock, S.D., and Ullman, G.L. (2005) Impacts of Using Dynamic Features to Display Messages on Changeable Message Signs, FHWA-HOP-05-069. Texas Transportation Institute, Houston, TX.
    
    
32. Charlton, S.G. (2006) “Conspicuity, Memorability, Comprehension, and Priming in Road Hazard Warning Signs.” Accident Analysis & Prevention, 38, p. 496–506.
    
    
33. Dudek, C.L., Schrock, S.D., Ullman, G.L., and Chrysler, S.T. (2006) “Flashing Message Features on Changeable Message Signs.” Transportation Research Record 1959, p. 122–129. Transportation Research Board, Washington, DC.
    
    
34. Dudek, C.L. and Ullman, G.L. (2002) “Flashing Messages, Flashing Lines, and Alternating One Line on Changeable Message Signs.” Transportation Research Record 1803, p. 94–101. Transportation Research Board, Washington, DC.
    
    
35. Dudek, C.L., Trout, N., Booth, S., and Ullman, G. (2002) Improved Dynamic Message Sign Messages and Operations, FHWA/TX-01/1882-2. Texas Transportation Institute, Houston, TX.
    
    
36. Dudek, C.L., Huchingson, R.D., Stockton, W.R., Koppa, R.J., Richards, S.H. and Mast, T.M. (1978) Human Factors Requirements for Real-Time Motorist Information Displays, Volume 1: Design Guide, FHWA-RD-78-5. FHWA, Washington, DC.
    
    
37. Dudek, C.L., Huchingson, R.D., Ratcliff, R.H., and Mayer, G.J. (1978) Human Factors Requirements for Real-Time Motorist Information Displays, Volume 7: Analysis of Driver Requirements for Intercity Trips. FHWA-RD-78-11. FHWA, Washington, DC.
    
    
38. Dudek, C.L., Huchingson, R.D., Koppa, R.J., and Edwards, M.L. Human Factors Requirements for Real-Time Motorist Information Displays, Volume 10: Human Factors Evaluation of Traffic State Descriptor Variables, FHWA-RD-78-14. FHWA, Washington, DC.
    
    
39. Carvell, J.D. and Whitson, R.H. (1978) Human Factors Requirements for Real-Time Motorist Information Displays. Volume 8: Analysis of Driver Requirements for Intracity Trips. FHWA-RD-78-12. FHWA, Washington, DC.
    
    
40. Carvell, J.D., Turner, J.M., and Dudek, C.L. (1978) Human Factors Requirements for Real-Time Motorist Information Displays. Volume 15: Freeway Incident Management Field Studies, FHWA-RD-78-19. FHWA, Washington, DC.
    
    
41. Ban, X., Li, Y., and Margulici, J.D. (2009) Optimal Use of Changeable Message Signs for Displaying Signs for Displaying Travel Times. University of California, Berkeley, CA.
    
    
42. Vercruyssen, M., Wagner, D.G., Hancock, P.A., Briggs, A., and Williams, G. (1997) Variable Message Signage Content: Survey Test of Meaning and Simulation Test of Usability, CTS 97-03. University of Minnesota Center of Transportation Studies, Minneapolis, MN.
    
    
43. Bloomfield, J.R. and Harder K.A. (2008) The Effectiveness and Safety of Traffic and Non-Traffic Related Messages Presented on Changeable Message Signs—Phase II, MN/RC 2008-27. Minnesota Department of Transportation, Minneapolis, MN.
    
    
44. Huchingson, R.D. and Dudek. C.L. (1983) “How to Abbreviate on Highway Signs.” Transportation Research Record 904, p. 1–4. Transportation Research Board, Washington, DC.
    
    
45. Hustad, M.W. and Dudek C.L. (1999) “Driver Understanding of Abbreviations on Changeable Message Signs in New Jersey.” Transportation Research Record 1689,
    p. 46–52. Transportation Research Board, Washington, DC.
    
    
46. AASHTO. (2011) Part I: Guidelines for the Selection of Supplemental Guide Signs for Traffic Generators Adjacent to Freeways, AASHTO, Washington, DC.
    
    
47. Boynton, R.M. (1971) Color Vision. Woodworth & Schlosberg's Experimental Psychology, 3d, Kling, J.W. and Riggs, L.A., Eds., Holt, Rinehart and Winston, Inc., New York, NY.
    
    
48. Bach, M. (1996) “The “Freiburg Visual Acuity Test”—Automatic Measurement of Visual Acuity.” Optometry and Vision Science, 73, p. 49–53.
    
    
49. Smart Eye Ab. (2010) Smart Eye Pro. http://www.smarteye.se/productseye-trackers/smart-eye-pro-2-6-cameras. Accessed December 4, 2012,
    
    
50. Inman, V.W. and Philips, B.H. (2013) Traffic Control Device Conspicuity. Technical Report, FHWA-HRT-13-044. FHWA, Washington, DC.
    
    
51. Bindemann, M., Burton, A.M., Hooge, I.T. Jenkins, R., and De Haan, E.H. (2005) “Faces Retain Attention.” Psychonomic Bulletin & Review, 12, p. 1,048–1,053.
    
    
52. Cerf, M., Frady, E.P., and Koch, C. (2009) “Faces and Text Attract Gaze Independent of the Task: Experimental Data and Computer Model.” Journal of Vision, 9, p. 11–15.
    
    
53. Lang, P.J., Bradley, M.M., Cuthbert, B.N. (no date) International Affective Picture System (IAPS). [Available online at http://csea.phhp.ufl.edu/media.html#topmedia.] Accessed December 6, 2012.
    
    
54. Stokes, M.E., Davis, C.S., and Koch, G.G. (2000) “Chapter 15: Generalized-Estimating-Equations.” Categorical Data Analysis Using the SAS System, 2d Ed., SAS Institute,
    Cary, NC.
    
    
55. Liang, Y., Lee, J.D., and Yekhshatyan, L. (2012) “How Dangerous Is Looking Away from the Road? Algorithms Predict Crash Risk from Glance Patterns in Naturalistic Driving. Human Factors.” The Journal of the Human Factors and Ergonomics Society, 54,
    p. 1,104–1,116.
    
    
56. National Highway Traffic Safety Administration. (2012) Visual-Manual NHTSA Driver Distraction Guidelines for in-Vehicle Electronic Devices, www.nhtsa.gov/staticfiles/rulemaking/pdf/Distraction_NPFG-02162012.pdf. Accessed February 4, 2013.
    
    
57. Stanislaw, H. and Todorov, N. (1999) “Calculation of Signal Detection Theory Measures.” Behavior Research Methods, Instruments & Computers, 31, p. 137–149.
    
    
58. Green, D.M., and Swets, J.A. (1966) Signal Detection Theory and Psychophysics, Wiley, New York, NY.
    
    
59. Mace, D.J., Hostetter, R.S., and Seguin, E.L. (1967) Information Requirements for Exiting at Interchange, 89211-F. Bureau of Public Roads, Washington, DC.
    
    
60. AASHO. (1958) American Association of State Highway Officials’ Manual for Signing and Pavement Marking of the National System of Interstate and Defense Highways, AASHO, Washington, DC.
    
    
61. Eyes Dx. MAPPS™. [Available online at http://www.eyesdx.com/products/mapps/.] Accessed February 8, 2011.
    
    
62. Agresti, A. (2002) Categorical Data Analysis, 2d Ed., John Wiley & Sons, Hoboken, NJ.
    
    
63. Dagnall, E., Katz, B., Bertola, M.A., and Shurbutt, J. (2013) Business Logo Signing: Evaluation of Number of Logo Panels on Specific Service Signs, TRB 92nd Annual Meeting, Transportation Research Board, Washington, DC.
    
    
64. Klauer, S.G., Dingus, T.A., Neale, V.L., Sudweeks, J.D., and Ramsey, D.J. (2006) The Impact of Driver Inattention on Near-Crash/Crash Risk: An Analysis Using the 100‑Car Naturalistic Driving Study Data, DOT HS 810 594. National Highway Transportation Safety Administration, Washington, DC.
    
    
65. SAE. (2007) Definitions and Experimental Measures Related to the Specification of Driver Visual Behavior Using Video Based Techniques: Surface Vehicle Recommended Practice. Society of Automotive Engineers, Warrendale, PA