In the summer of 1978, welding crews work at night, in secret, on the diagonal joints of New York's Citicorp Center — today 601 Lexington Avenue — while the building stays occupied during the day and the city knows nothing about it. Engineer William LeMessurier had discovered, after construction was complete, that the bolted connections of the bracing diagonals — substituted during execution for the welds originally specified, to save time and cost on site — weren't able to withstand "quartering winds" (winds that hit the building at a corner rather than perpendicular to the façades), a load case the original analysis hadn't fully considered. With a hurricane forecast and an estimated probability of failure that was far from negligible, the decision was to reinforce every critical joint in a rush, in secret. No profile was undersized: the problem was the joints.
This episode, now a textbook case in structural engineering education, sums up better than any manual a principle that holds truer for steel than for any other building material: the connection, not the profile, is where a steel structure proves — or disproves — its own reliability.
Bolted versus welded: two site philosophies
Bolted connections join elements with high-strength bolts, typically grade 8.8 or 10.9, working in shear (the bolt itself resists the transverse load) or in friction (the clamping preload presses the plates against each other and the force transmits by friction, with the bolt shank not stressed in shear under service conditions). The main advantage is speed and site-verifiable quality: a bolt is tightened with a torque wrench, the result can be checked immediately, it doesn't require labor as specialized as a certified welder, and it can be carried out in any weather.
Welded connections join elements by directly fusing the metal, in the workshop or on site, with full-penetration or fillet welds. The advantage is near-perfect structural continuity — a properly executed weld behaves as a monolithic material — and the absence of visible connecting elements, preferred when aesthetics call for clean surfaces. The disadvantage is sensitivity to execution quality: a defective weld (porosity, lack of fusion, hydrogen cold cracking) is a hidden flaw that requires costly non-destructive testing — ultrasound, radiography, magnetic particles — to detect, and is far more affected by site conditions (wind, humidity, temperature) than by workshop ones.
Current practice, in most contemporary steel structures, is a compromise: welding in the workshop, where conditions are controlled and quality checks systematic, and bolting on site, where speed and immediate verifiability matter more. This is exactly the logic that, in the Citicorp case, was applied to the wrong joints.
Capacity design: the hierarchy of strengths
In seismic zones, the choice of connection type isn't just a site matter: it's the basis of an entire design method called "capacity design," or the hierarchy of strengths. The principle, now codified in the Eurocodes and seismic codes worldwide, is that a structure must be designed so that, under a severe earthquake, failure occurs at chosen, controlled points — typically the beams, which deform plastically in a ductile way, dissipating energy — while columns and connections remain overstrength and don't collapse in a brittle manner. It's the principle known as "weak beam, strong column": an element is allowed to give way, but it must be the right one, failing in the right way.
The 1994 Northridge earthquake in California put this principle to a dramatic test: numerous steel moment-resisting frames, considered among the most reliable seismic-resistant systems in existence, suffered unexpected brittle fractures precisely at the welds connecting beam flanges to columns — points the hierarchy-of-strengths theory was supposed to protect. The post-earthquake investigation, coordinated by the federal SAC Joint Venture project, attributed the problem not to the capacity-design principle itself, but to specific weld construction details (notches, inadequate filler metal, geometries that concentrated stresses). The result was a thorough overhaul of prequalified connection details — including the reduced beam section (RBS, or "dogbone"), which deliberately weakens the beam at a point away from the joint to ensure the plastic hinge forms there, not in the weld. Even the system meant to be foolproof, in short, turned out to be only as reliable as its worst joint.
Bracing systems
A steel frame on its own resists horizontal forces (wind, earthquake) poorly if its beam-column joints are simple pins: it needs a bracing system. There are three main options. Concentric (diagonal) bracing, with diagonals arranged in an X, an inverted V (chevron) or a K, transfers horizontal forces through pure axial tension/compression in the members — it's the stiffest and cheapest system, but it introduces diagonal elements that constrain the layout of doors, windows and circulation. The moment-resisting frame (MRF), where beam-column connections also transmit bending moment, needs no diagonals and leaves bays completely free — at the cost of greater lateral deformability and connections that are far more expensive to execute and inspect, exactly the ones that came under scrutiny after Northridge. A third family, the eccentrically braced frame (EBF), inserts a short "sacrificial" beam segment (link) between the diagonal and the joint: in a severe earthquake, that segment deforms plastically in a ductile, predictable way, combining the stiffness of concentric bracing with a dissipative capacity that pure X-bracing doesn't have.
In very tall buildings, bracing shifts scale: outrigger systems connect the central core (where elevators and stairs are usually concentrated) to the perimeter columns through one or more truss systems as deep as an entire floor, often paired with a belt truss that distributes the effect to all the perimeter columns. The result is a drastic increase in the building's overall torsional and flexural stiffness, essential above a certain height: both Shanghai's Jin Mao Tower and Taipei 101 use outrigger systems to keep top displacements within occupant comfort limits.
The bracing that moves: tuned mass dampers
There's a fourth way to control the horizontal sway of a tall building, one that doesn't stiffen the structure but goes along with it: the tuned mass damper (TMD), an enormous mass suspended at the top of the building on cables and dampers, calibrated (hence "tuned") to oscillate out of phase with the wind-induced motion, dissipating its energy. The same 1978 Citicorp Center was among the first examples of this in the world: a mass of about 400 tonnes installed on the 63rd floor specifically to reduce the sway perceived by occupants on high-wind days — a device conceived for comfort, not structural safety, but one that same year's bolt crisis suddenly made central to safety too. The most spectacular case today remains the 660-tonne gold sphere suspended between the 87th and 92nd floors of Taipei 101 (2004), visible to the public and itself now a tourist attraction; Boston's John Hancock Tower installed two 300-tonne masses in 1977 for the same reason. None of these devices replaces structural bracing: it acts in parallel, on comfort and on material fatigue from repeated oscillation, not on ultimate strength.
When the joint becomes façade
If Mies van der Rohe had made the steel frame elegant and quiet, the high-tech firms of the 1970s and '80s turned connections and bracing into a full-blown, declared architectural language. Paris's Centre Pompidou (1977, Piano, Rogers, engineer Peter Rice) moves the structure entirely outside the envelope, with its famous cast-iron spherical joints — the "gerberettes" — designed specifically for that building. The Lloyd's of London building (1986, Rogers) turns the entire load-bearing structure into an exposed exoskeleton. Chicago's John Hancock Center (1969, SOM, with engineer Fazlur Khan) displays on its façade the huge bracing X's of its "trussed tube," a system efficient enough to drastically cut the amount of steel needed compared with a conventional frame — and photogenic enough to become the building's very image. Hong Kong's Bank of China Tower (1990, I.M. Pei) uses an external diagonal lattice to transfer loads down to just four corner columns. In all these cases, the connection and the bracing stop being a detail to hide behind a suspended ceiling: they become the central subject of the design.