A bearing passes the superstructure's reaction to the pier while letting the deck rotate and, where required, translate. Every bearing layout is an articulation scheme: which supports are fixed, which are guided, which are free, chosen so that temperature can move the deck without fighting itself and seismic or braking force has a defined path to the ground.

• Elastomeric bearings, layers of rubber bonded to steel shims. Cheap, maintenance-free, accommodate rotation and moderate translation by shear deformation. The default for spans under about 50 m and reactions to a few meganewtons.
• Pot bearings, a rubber disc confined in a steel pot, taking large loads and rotations; sliding PTFE-on-stainless surfaces add translation.
• Spherical and disc bearings, for large rotations and very large loads, long-span and railway bridges.
• Rocker and roller bearings, historical steel bearings that seize with corrosion; the locked bearing is a classic cause of unintended forces and cracked piers.

Design a bearing to be replaced: leave jacking points and clearance, because it will be.

Joints let the deck breathe and are the single most damaging component on a bridge: when they leak, salt water runs onto the girder ends, bearings and pier caps below, and that is where most of the deterioration in a cold-climate bridge is found. The first design decision is therefore how many joints can be eliminated, by continuity, by integral abutments, by link slabs.

Where a joint is unavoidable, its type follows the movement: asphaltic plug joints for a few tens of millimetres, strip seals (a rubber gland between steel extrusions) to about 100 mm, modular joints with several seals for movements of hundreds of millimetres on long bridges, and finger joints where large movements must carry heavy traffic. Movement is computed from temperature range, creep and shrinkage, and, in seismic zones, from the design displacement; installation temperature sets the gap.

Water is the enemy, and drainage is the defence that costs least and is neglected most. A deck is crowned or cross-sloped at about 2% and graded longitudinally so that water reaches inlets (scuppers) sized for the design storm; downspouts then carry it clear of girders, bearings, piers and the ground below the abutments. Deck ends drain away from the joint; girder bottom flanges have drip grooves; box girders have vents and drains in every cell.

Inspections consistently find the same faults: scuppers blocked with debris, downspouts discharging onto pier caps, and joints leaking, each one a maintenance item that turns into a structural one.

Two design rules cover most of it. Inlets every 20–30 m on a 2% longitudinal grade, closer where the grade flattens; and no free discharge onto anything below: piers, girders, bearings, railways, roads or navigable water all need piped drainage to the ground. Inspectors should be able to reach every scupper and every cleanout.

Place an inlet just upstream of every expansion joint and at the low point of every sag curve. Keep the pipework generous: AASHTO LRFD Article 2.6.6 asks for downspouts of at least 150 mm, or 200 mm where ice is expected, and every bend needs a cleanout. FHWA's HEC-21, Design of Bridge Deck Drainage, is the standard reference for inlet interception and spacing.

Barriers keep vehicles on the bridge. Modern barriers are not designed by calculation but by crash test: MASH (formerly NCHRP 350) test levels TL-1 to TL-6 define the vehicle, speed and angle, and a barrier is qualified at a level, TL-4 for most highways, TL-5 where heavy trucks are frequent. The concrete F-shape and single-slope profiles, and steel post-and-rail systems, are the common qualified shapes.

What the designer calculates is the deck overhang that carries the barrier: AASHTO Section 13 gives the transverse and longitudinal collision loads that the overhang and its reinforcement must resist, and, as Chapter 6 noted, they usually govern the overhang design. Pedestrian and bicycle railings have their own height and load rules.

Bridges carry more than traffic: water mains, gas, power, telecoms, lighting and signs. Each is a permanent load (DW), some are dynamic (a bursting main is a real event), and all need access for the utility's own maintenance without closing the bridge or damaging it. Attachments are the fatigue details nobody designed; a sign bracket welded to a girder flange is a Category E' detail waiting for a crack.

Inspection access, walkways in box girders, ladders in towers, catwalks under decks, gantries on long spans, is designed in from the start on well-run projects and improvised for decades on the rest.

Utilities also decide inspection access and, in the worst case, the failure mode: fires fed by a utility corridor have destroyed decks, and gas mains are kept outside enclosed boxes for that reason.

Lighting columns and sign gantries on a bridge are fatigue structures twice over: wind excites them (vortex shedding, natural gusts, and the gusts of trucks passing under a sign), and they stand on a deck that vibrates under traffic. They are designed to the AASHTO LRFD Specifications for Structural Supports for Highway Signs, Luminaires, and Traffic Signals, placed near piers where deck vibration is smallest, and given base details chosen for fatigue rather than convenience. Tall pylons carry aviation obstruction lights and navigable spans carry channel lights, each needing power and a maintenance route that will last.

Inspection access is best treated as a design deliverable with its own drawing, showing how an inspector reaches every bearing, every joint underside, every stay anchorage and every cell of every box. Box girders need hatches at the ends and openings through every diaphragm, lighting, power and ventilation, and a rescue plan, because they are confined spaces. Long spans justify permanent under-deck inspection gantries running on rails; shorter ones need at least anchor points and a clear path for an under-bridge inspection vehicle. An element that cannot be reached will not be inspected, and the element that is not inspected is the one that fails.

Rubber is nearly incompressible but very soft in shear. A plain pad squeezed between two surfaces bulges sideways and carries little; bond thin steel shims into it and each layer is restrained from bulging, so the pad becomes stiff in compression while staying soft in shear and rotation. The governing number is the shape factor of a layer, the loaded area divided by the area free to bulge: S = LW / [2h_ri(L + W)] for a rectangular pad of plan dimensions L × W and layer thickness h_ri. Compressive stiffness rises roughly with S², so thin layers make stiff bearings; shear stiffness, GA / h_rt, depends only on the total rubber thickness h_rt.

AASHTO LRFD offers Method A (Article 14.7.6), simple and conservative, and Method B (Article 14.7.5), which permits higher stresses in exchange for more testing and a combined shear-strain check; EN 1337-3 adds the shear strains from compression, translation and rotation and compares the sum with a single limit. The shear modulus G of bearing elastomer lies between about 0.55 and 1.2 MPa, with 0.9 MPa a common nominal value; elastomer stiffens in the cold, so the compound is graded for the site's low-temperature zone (AASHTO M 251).

The Method A checks for a steel-reinforced pad, in words:

• Compression. Average service stress σ_s ≤ 1.25 GS and ≤ 8.6 MPa (1.25 ksi).
• Shear. Total rubber thickness h_rt ≥ 2Δ_s, which keeps the shear strain from movement below 50%.
• Rotation. The pad must take the girder end rotation, plus an allowance of 0.005 rad for construction tolerance, without lifting off at one edge or overstraining the other.
• Stability. Total thickness not more than L/3 or W/3, so the pad cannot roll over.
• Anchorage. If the factored shear force exceeds one-fifth of the minimum permanent vertical load, friction cannot be relied on and the pad must be secured.

The last line is the one most often forgotten. An elastomeric bearing is a spring, not a slider: six girders on such pads push about 260 kN into the abutment or pier each time the deck shortens, and a substructure designed as if its bearings were frictionless will be under-reinforced. The other recurring faults are pads that walk out of position under low permanent load, uneven seating that loads one edge, and the belief that a pad needs no inspection.

Above a few meganewtons, or where rotations and movements exceed what a pad can take, bearings become mechanical. A pot bearing confines an elastomer disc in a steel cylinder under a close-fitting piston; confined on all sides, the rubber behaves almost like a fluid, carrying high pressure while letting the piston tilt. Its weak point is the internal seal ring that stops the rubber extruding past the piston, which wears under many rotation cycles. A spherical bearing rotates by sliding a convex steel calotte in a concave seat lined with PTFE or a modern polymer, giving large rotations with little resistance. Either type becomes a sliding bearing by adding a flat sliding surface: dimpled, lubricated PTFE against mirror-polished stainless steel. In Europe the family is covered by EN 1337 (Part 2 sliding elements, Part 3 elastomeric, Part 5 pot, Part 7 spherical and cylindrical PTFE, Part 8 guided and restraint bearings); in the United States by AASHTO LRFD Section 14.

Each bearing is classed by what it restrains: fixed (rotation only), guided (slides in one direction and is restrained across it) or free (slides in any direction). The pattern of these choices across the bridge is its articulation, and good articulation follows a handful of rules.

1. One fixed point per continuous deck. Thermal movement radiates from it and the joints sit at the ends. A fixed point near the middle halves the movement at each joint; a fixed point at the stiffest support gives braking and seismic forces the shortest path to the ground. Two fixed supports far apart lock the deck between them and generate large forces at every change of temperature.
2. Guides point at the fixed point. On a curved deck the expansion runs along the straight line from the fixed point, not along the curve, so guided bearings are aligned on that line; aligned on the tangent, they fight the deck.
3. One transverse restraint per support line. A wide deck also expands sideways: guide one bearing per support, usually the one nearest the centreline, and leave the others free.
4. Count the friction. Free bearings are not frictionless. At 3% friction, a 30,000 kN deck delivers up to 900 kN to its fixed point as it slides, and because the bearings on the two sides of the fixed point do not slide equally freely, the fixed support is designed for the unbalanced friction, not for zero.
5. Avoid uplift. Short end spans, sharp curves and skews can lift a bearing off its seat under live load. Counterweight the end, change the span arrangement or use a tension bearing; never assume it away.

Two more rules belong to the life of the bridge. Every bearing will be replaced, so the drawings show jacking points designed for the full reaction, the lift needed (a few millimetres) and the clearance to withdraw the old bearing. And movable bearings are pre-set at installation for the measured deck temperature, with their transport clamps removed only once the deck sits on them; a clamp left in place turns a free bearing into a fixed one, and has cracked more than one pier.

How hard an earthquake shakes a bridge depends mostly on its natural period, and a short, stiff bridge on squat piers, with a period of a few tenths of a second, sits where most ground motions are strongest. Seismic isolation places a soft, damped layer between the deck and the substructure. It lengthens the period to 2–3 s or more, where the spectral acceleration is much lower, and adds damping that lowers it further. The price is displacement: the deck moves a few hundred millimetres on its isolators, and everything that crosses the isolation plane (joints, seats, utilities, the approach slab) must accommodate that movement.

Three families do most of the work:

• Lead-rubber bearings (LRB): a laminated elastomeric bearing with a lead core that yields in shear at about 10 MPa. They are stiff under wind and braking, and in an earthquake they follow a bilinear, energy-absorbing loop.
• High-damping rubber bearings (HDRB): a rubber compound with built-in damping of roughly 10–20% of critical, without a core.
• Friction pendulum bearings: an articulated slider on a concave stainless steel surface of radius R. The deck rises slightly as it slides, so gravity pulls it back towards the centre, while friction dissipates energy. The pendulum period is T = 2π√(R/g), independent of the mass carried.

Isolation is designed to the AASHTO Guide Specifications for Seismic Isolation Design, or to EN 1998-2 Section 7 with the devices to EN 15129. Two ideas set it apart from ordinary bearing design. First, device properties are uncertain and change with temperature, ageing, wear and, for lead cores, heating during the shaking itself, so the analysis is run twice: with upper-bound properties for the forces in the substructure and lower-bound properties for the displacements. Second, the devices are qualified by full-scale prototype tests at the design displacement, and production bearings are tested before they leave the factory.

Isolation suits stiff bridges on firm ground in high-seismic regions, and existing bridges whose piers cannot be made ductile. It helps less where the bridge is already flexible (tall piers already have long periods), on soft soil whose ground motion is rich in long periods, and near an active fault, where a velocity pulse can demand displacements far beyond the design value. That lesson came from the Bolu viaduct in Türkiye, a 2.3 km viaduct of 40 m prestressed concrete girder spans on pot bearings with steel energy dissipation units, nearly complete when the Düzce earthquake struck on 12 November 1999. The fault rupture crossed the structure; deck movements exceeded the design limits, many of the dissipation devices were damaged, and some girders were left with only a few centimetres of seat. The viaduct was repaired with friction pendulum bearings, chosen because they could accommodate a large permanent fault offset and still dissipate a great deal of energy.

A joint is sized for the movement of the deck end between its extreme positions, counted from the day the joint is set. The contributions, roughly in order of size:

• Uniform temperature. Δ = αLΔT, where L is the distance from the fixed point, α is about 10–11 × 10⁻⁶ /°C for concrete and 12 × 10⁻⁶ /°C for steel. AASHTO LRFD Article 3.12.2 gives the design temperatures either from a simple table (Procedure A) or from contour maps of maximum and minimum design temperature (Procedure B). EN 1991-1-5 derives the effective bridge temperature from the shade air temperatures of the site, and for bearings and expansion joints (6.1.3.3) widens the expansion and contraction ranges by 20 °C, or by 10 °C when the setting temperature is specified.
• Shrinkage and creep of concrete decks, which only shorten the deck and continue for years after the joint is installed.
• End rotation. A girder end rotating about its bearing moves the deck surface horizontally by the rotation times the height of the surface above the bearing: 0.005 rad on a girder 2.5 m deep is 12 mm.
• Braking and seismic displacement: the first small, the second provided for in full or by a joint designed to be damaged and repaired.

A single strip seal cannot take 163 mm, and a two-cell modular joint at 80 mm per cell has no margin left for tolerance, so the choice is a three-cell modular joint or a finger joint. The same numbers fix the setting gap. The joint must be able to close 48 mm and open 115 mm from the day it is set, so a three-cell joint with a range of 0–240 mm is set near the closed end of its range, at roughly 60–70 mm in total, not at mid-range. If the deck is at 22 °C on the day instead of 30 °C, it is 8 × 1.6 ≈ 13 mm shorter, and the gap is set 13 mm wider; joint drawings carry a presetting table by deck temperature for exactly this reason.

Two habits reduce trouble. Install joints as late as the programme allows, so that part of the shrinkage and creep has already happened. And measure the temperature of the deck, not the air, on the day: the commonest joint fault is a gap preset for an assumed temperature. Skew adds a second check, since the movement resolves into opening across the joint and shear along it, and strip seals and modular joints have limits on both.

A modular joint divides a large movement among several small gaps. Steel centre beams run across the roadway, separated by elastomeric strip seals of about 75–80 mm movement each. They rest on support bars spanning the gap, which slide on bearings inside boxes cast into the deck end and the abutment, while springs or a linkage keep the gaps equal. In a multiple support bar joint each centre beam is welded to its own support bars; in a single support bar (swivel-joist) joint each support bar carries all the centre beams through sliding yokes, so one bar does the work of many, at the cost of more moving parts.

Every axle of every truck strikes every centre beam, as a sharp impact with substantial dynamic amplification, repeated hundreds of millions of times over the life of a joint on a busy road. Early modular joints cracked by fatigue at the welded connection between centre beam and support bar and at centre beam splices, some within a few years of opening. The response in the United States began with NCHRP Report 402 (1997) on fatigue design, and AASHTO LRFD Article 14.5 now calls for modular joints to be designed for infinite fatigue life, with stress ranges below the constant-amplitude fatigue threshold, and to be qualified by testing. In Europe, joints are assessed against the requirements first set out in ETAG 032.

Noise, a rhythmic clatter as each axle crosses each beam, is reduced in urban areas by surface plates with sinusoidal or rhomboid edges, so that a tyre never drops into a straight gap. Grit packs the seals; broken control springs let the gaps become unequal until one seal tears; the sliding bearings inside the boxes wear. A modular joint is a machine, and it needs a chamber or access from below so that its underside can be inspected and its parts replaced.

A finger joint bridges the gap with interlocking steel teeth, cantilevered from one or both sides or supported on the far side. It carries heavy traffic smoothly over large movements with few moving parts, but it is open: water and grit fall through, so it needs a drainage trough beneath, which in turn needs cleaning. The gaps run in the direction of travel, which matters for bicycles and motorcycles and limits the gap width; the finger tips must not stand proud of each other when the deck end rotates; and the anchor bolts and finger roots are fatigue details in their own right.

Installation decides much of a joint's life. The joint sits in a recess at the deck end, flush with the surfacing to within a few millimetres, preset from the measured deck temperature, anchored with loops or studs into a high-strength concrete or elastomeric concrete nosing, and given a drainage path for the water that will eventually get past the seal. Most early joint failures are installation faults: a recess poured with voids under the support boxes, a joint set proud of the road so that tyres and ploughs strike it, or a seal installed twisted.

Joints and the bearings below them cause most of the maintenance on short and medium bridges (section 9.2), so the best joint is none. Three techniques remove them.

Integral abutments. The deck is cast monolithically into the end walls, which stand on a single row of flexible piles, typically steel H-piles or small-diameter concrete piles. The whole bridge expands and contracts as one frame, the piles bending to follow the abutment; to make them more flexible they are often driven through pre-bored holes or sleeves filled with loose material for the top few metres. An approach slab tied to the abutment slides on the fill and moves the joint, now a simple seal or asphaltic plug, to the far end of the slab, where a leak drips onto the embankment instead of a bearing seat. UK practice (CD 350) expects integral construction to be considered for bridges up to 60 m long with skews up to 30°; many US states build integral bridges well beyond 100 m.

The price is soil-structure interaction. Each summer the abutment pushes into the backfill and each winter it pulls away; the backfill slumps a little into the gap and is compacted again the next summer, so the earth pressure builds up over the years towards passive pressure, a process known as ratcheting. In the UK, PD 6694-1 gives the design earth pressure as a function of the wall's thermal movement relative to its height. On short bridges the movements stay modest:

Semi-integral abutments cast the end diaphragm, or an end screen wall, monolithically with the deck but sit the girders on bearings on a conventional abutment. The deck end moves against the backfill as in an integral bridge, but the foundations do not have to bend, which suits rock near the surface, tall abutments and stiff foundations. The bearings remain, but out of the water path, and the joint is gone.

Link slabs remove the joints over the piers of a chain of simply supported girder spans, and are among the commonest joint-elimination retrofits. The deck slab is made continuous across the pier while the girders below remain simply supported. Over a length each side of the pier, about 5% of each span being a commonly used value, the slab is debonded from the girder tops so that it is flexible enough to follow the girder end rotations, and it is reinforced to control the cracks those rotations cause. Ductile materials such as engineered cementitious composites and UHPC have been used to make the link itself crack-tolerant.

Skew, curvature and length still need judgement, but on most short and medium bridges the first question is not which joint to use, but why there should be one at all.

The two crash-test systems most engineers meet are the American Manual for Assessing Safety Hardware (MASH) and the European EN 1317. MASH replaced NCHRP Report 350 in the United States, and bridge rails on new National Highway System projects let after 2019 must be MASH-tested.

AASHTO LRFD Section 13 and its Appendix A13 turn a test level into design forces for the barrier and the deck. Table A13.2-1 still reflects the NCHRP 350 impact conditions: for TL-4 a transverse force F_t of 240 kN (54 kips) distributed over 1.07 m and applied 810 mm above the deck, for TL-5 551 kN (124 kips) over 2.44 m at 1,070 mm. Research for the heavier MASH vehicles has found larger demands: for TL-4, a minimum rail height of 915 mm (36 in) and transverse design forces of roughly 310–355 kN (70–80 kips), depending on the rail height.

For a concrete barrier, Appendix A13 uses yield-line analysis: the impact spreads over a critical length L_c of wall, and the wall's capacity R_w follows from its flexural strength about vertical and horizontal axes and that of any cap beam. R_w must exceed F_t. The deck overhang is then designed for the barrier's actual capacity, not for F_t: the moment at the barrier base and the tension R_w / (L_c + 2H) that the barrier delivers to the slab, so that a severe impact breaks the barrier and leaves the deck intact. Eurocode applies the same principle: EN 1991-2 gives transverse collision forces for classes A to D (100 to 600 kN), and asks that the structure supporting a parapet be designed for at least 1.25 times the parapet's characteristic local resistance.

Every level from N2 upwards is also tested with a 900 kg car at 100 km/h and 20°, because a barrier strong enough to stop a lorry must not be so stiff that it kills the occupants of a small car. That is measured by the acceleration severity index (ASI class A ≤ 1.0, B ≤ 1.4, C ≤ 1.9); the barrier's deflection is measured by its working width class, from W1 (0.6 m) to W8 (3.5 m), which must fit the space behind it, on a bridge the distance to the edge or to the next hazard.

Selection follows the road and what lies beyond the edge; the very high levels are kept for decks over railways, busy roads and populated areas, where a vehicle going over the edge would cause a second disaster. Two details deserve as much care as the barrier: the transition to the approach guardrail, which must be crash-tested as a system because an abrupt change in stiffness snags vehicles; and pedestrian and bicycle railings, at least 1,070 mm (42 in) high under AASHTO LRFD Articles 13.8 and 13.9.