Washington State’s design manual states the awkward part in its opening paragraph on traffic barriers. In some cases, it says, installing a barrier may result in more crashes, because the barrier is an object that can be struck. The manual goes further a paragraph later: traffic barriers do not prevent crashes or injuries from occurring, they often lower the potential severity of crash outcomes, and consequently they should not be used unless a reduced crash severity potential is likely.

That is the whole discipline in three sentences. A guardrail is a hazard an agency installs on purpose, on the judgment that hitting it will hurt less than hitting whatever is behind it. Every argument about roadside design is an argument about whether that judgment holds at a particular location.

Shielding is the fourth choice, not the first

The order of preference is explicit and it is not what a driver looking at a rail would guess. Washington’s roadside chapter directs designers to consider mitigation in this sequence: remove the object, relocate it, redesign it with breakaway features or make it traversable, shield it with a traffic barrier, or delineate it. Barrier is fourth of five. Delineation, the option of last resort, requires its own documented design analysis.

The factors that decide among them are equally plain: crash severity potential, maintenance needs, and cost across both the initial build and the life cycle. Maintenance is not a footnote here. A less rigid system absorbs more damage per impact, which means more repair visits, which means crews working beside live traffic more often. The barrier that performs best in a crash test can be the one that generates the most exposure over thirty years, and the manual tells designers to consult the area maintenance superintendent before choosing.

Colorado states the purpose from the other direction. Highway guardrail exists to help prevent an errant vehicle from crashing into roadside obstacles or into oncoming traffic, and the things it is there to shield are named: median structures, bridge piers, culvert headwalls, non-traversable ditches, trees. Those should be removed, redesigned, or shielded, in that order. The rail is the fallback for hazards that cannot be got rid of.

What the clear-zone table actually asks for

Clear zone is the roadside area kept traversable so a driver or cyclist who leaves the lane can recover. It is measured from the edge of the through traveled way, in both directions on a two-way road, and it is a target value rather than a guarantee. Washington’s manual concedes outright that it is not possible to provide a clear zone free of objects at all locations and under all circumstances.

The width comes from a lookup on posted speed, side slope and traffic volume, and the spread is wider than most people expect. At 35 mph or less the design clear zone is 10 feet regardless of slope or volume. At 70 mph with more than 6,000 vehicles a day, a 4H:1V fill slope calls for 54 feet, while a cut section with a 3H:1V backslope at the same speed and volume calls for 24. The same road, the same traffic, and a factor-of-two difference in how much land the roadside design consumes, driven by which way the ground falls.

Slopes between 4H:1V and 3H:1V create the interesting case. They count as traversable if free of fixed objects, and they are also classified as nonrecoverable, meaning a vehicle can cross them without regaining control. The table gives no clear zone distance for them at all. Instead the horizontal length of the slope is added to the clear zone to produce a recovery area, which in the manual’s worked example comes to 30 feet at 45 mph. Roadside width is a right-of-way problem before it is a hardware problem.

Why the crash-test standard changed

Barriers are tested under standardized conditions, because it is impossible to replicate the variations in real highway geometry. Those conditions lived in NCHRP Reports 230 and then 350, and now live in the AASHTO Manual for Assessing Safety Hardware.

Colorado gives the reason for the change without diplomatic padding. Today’s passenger vehicles have become higher, larger and heavier than the vehicles crash tested in the 1990s under NCHRP Report 350, and crash testing has shown that some products tested under the older protocol failed to meet current vehicle conditions. The fleet outgrew the test.

Texas describes the resulting structure. MASH test level 2 covers low-speed roadways at 45 mph or less; test level 3 covers high-speed roadways at 50 mph or greater; test level 4 adds a 22,000-pound delivery truck to the level 3 matrix. The standard vehicles are a 2,420-pound small car and a 5,000-pound large pickup, and the key change from the older protocol was increasing the size and height of the tested pickup to reflect the fleet and better simulate an SUV. A device becomes eligible for federal-aid reimbursement once a state can certify that an accredited laboratory tested it and it met the criteria.

The transition dates are specific. FHWA stopped issuing eligibility letters under NCHRP 350 testing on December 31, 2015, and from January 1, 2016 new devices needed MASH testing. Category deadlines followed: W-beam barriers and cast-in-place concrete barriers at the end of 2017, W-beam end terminals at the end of June 2018, crash cushions at the end of 2018, and everything else at the end of 2019.

What happens to the hardware already in the ground is where two state manuals read differently, and the difference is instructive rather than a contradiction. Texas states that hardware accepted under NCHRP Report 350 does not need retesting, may remain in place, and may continue to be manufactured and installed. Washington requires its projects to replace pre-NCHRP 350 hardware, excepting breakaway cable terminals with their own replacement program, and lets improvement and preservation projects leave NCHRP 350 compliant hardware alone while it stays serviceable. Both are consistent with a schedule that governs new installations rather than the standing inventory.

The height nobody adjusts

The most instructive detail in this field concerns a number that changes without anyone touching the guardrail. Colorado’s guidance sets 31 inches as the target height for new installations of the Midwest Guardrail System, up from 28 inches, because MASH testing showed better containment and redirection of vehicles with a higher centre of gravity such as pickups and SUVs. It also sets a floor: after a pavement overlay, the rail’s minimum height is 26.5 inches, a figure it attributes to the AASHTO Roadside Design Guide.

Suppose a resurfacing project raises the pavement by two inches. The rail has not moved, the posts have not moved, and the effective height has dropped two inches toward a threshold that triggers remedial work. Colorado’s answer is to move the guardrail and blocks up three inches on the post to the pre-drilled bolt hole, or to extract and re-drive the posts at the new elevation if the barrier is being relocated anyway. This is why roadside safety turns up as a line item on paving contracts.

The transition between heights is engineered rather than improvised, at 25 feet of W-beam panels to step from 28 inches to 31.

What a barrier costs somewhere else

Barrier selection has consequences that never appear in a crash-severity calculation. Washington’s manual requires an environmental review before concrete barrier goes in near stormwater facilities or wetlands, because it adds impervious surface and can force reconstruction of an existing detention system. The runoff consequences are the subject of highway stormwater runoff.

The wildlife finding is sharper. When animals meet a physical barrier they cannot easily see beyond or cross, such as concrete barrier, they often stop or move parallel to it, increasing their time on the highway and their exposure. A device installed to reduce crash severity for vehicle occupants can raise the collision risk for the species crossing that corridor, which is one reason wildlife crossings are built as structures rather than as gaps in a rail. Beam guardrail and concrete barrier also catch drifting snow, add shoulder widening and earthwork, and can cut sight distance, all of which the manual lists as selection criteria alongside crash performance, the evidence base that highway safety coverage tracks treatment by treatment.

The guidance that is not a federal standard

One last thing is worth checking before treating any of this as binding. The roadway design documents adopted by reference in 23 CFR 625.4 are the AASHTO Green Book, seventh edition, and the Interstate design policy, followed by a list of bridge, structural support and welding specifications. The AASHTO Roadside Design Guide is not among them.

So the clear-zone tables, the mitigation hierarchy and the barrier warrant that shape every American roadside are state policy built on a national guide that federal design standards do not incorporate. That arrangement gives agencies room to fit the roadside to local terrain and right-of-way, and it also means a driver’s protection on a given mile depends on which state manual was open when the plans were drawn. The clear zone at 70 mph is 54 feet in Washington because Washington decided so.