Twenty-six years ago this morning, Tom Sabol was jolted awake by the thrashing of an earthquake. He lived some 20 miles from the epicenter in suburban Northridge. But he still remembers power lines sparking and fizzling in his West LA neighborhood.

At the time, Sabol, who has a doctorate in structural and earthquake engineering, was working on the Getty Center, the largest construction project underway in Los Angeles in the mid-1990s. But amid the tumult, he wasn’t thinking about the Getty. He was so confident in the steel building skeletons he had helped engineer that it didn’t cross his mind that they might yield to the shaking.

“We thought the Getty had been designed for a level of earthquake demand greater than the minimum requirements in the building code—by a substantial amount,” says Sabol, a principal with engineering firm Englekirk Institutional.

That confidence was shattered about a week later when the Getty’s contractor discovered cracks in the welding in at least one of the steel-frame buildings being constructed. According to a February 1994 article in the Los Angeles Times, six of 80 joints were damaged, and one cracked in two. One steel column cracked.

The Northridge earthquake showed the vulnerability in the welded and bolted connections of what are called moment frames. Until then, they had been considered seismic-proof. And this wasn’t just at the Getty. Engineers considered moment frames to be “among the most ductile systems contained in the building code.” They had been widely used in steel buildings since the 1960s.

On February 27, 1994, about a month after Northridge rocked Southern California, the LA Times reported that welds had fractured and columns had cracked in buildings across the region, from West LA to Valencia. Some buildings had only been constructed within the last five years—these were modern buildings that had been “thought to be safest.”

“This wasn’t supposed to happen,” Nestor Iwankiw, then-president of the American Institute of Steel Construction, told the newspaper. “This is the first time this type of failure has been seen.”

The tally of steel-framed buildings damaged in the earthquake would ultimately climb to 100.

In the weeks and months that followed the Northridge earthquake, the Getty assembled a team of researchers, industry specialists, and steel fabricators to suss out where the structural weaknesses were—and how to fix them. Relatively quickly, they found a solution for the Getty connections, and their work set the stage for a larger research effort sponsored by Federal Emergency Management Agency. Because of the Getty’s quick action, building codes and inspection practices changed across the industry to require stronger materials, higher-quality welds, and more rigorous testing for every new type of connection.

In Los Angeles, the city ordered inspections of 239 welded steel moment frame buildings in West LA and the San Fernando Valley, and in buildings where damage was found, the city required owners to make repairs.

But an untold number of welded steel moment frame buildings constructed prior to 1994 that are outside of those regions have yet to be inspected, repaired, or retrofitted in the city of Los Angeles.

It’s almost certain that an earthquake larger than Northridge will strike Southern California in the next 30 years. In Los Angeles, city officials have ordered thousands of property owners to retrofit non-ductile concrete buildings and “soft-story” buildings that that proved especially vulnerable in 1994. But the city has yet to require seismic retrofits of steel buildings outside West LA and the Valley, areas that experienced the most ground acceleration during Northridge. It’s unknown, for example, how many steel skyscrapers in Downtown Los Angeles have cracks that haven’t been discovered.

Among those who rushed to Los Angeles in the days and weeks following Northridge was Mike Engelhardt, a researcher and engineering professor at the University of Texas at Austin.

When he got to the Getty, he observed cracks in welded connections that were large enough to slide his business card through. “It should have been solid steel—with no cracks there whatsoever,” he says. “Even in a very strong earthquake, welds should not fracture.”

He was surprised. But not that surprised. One month before, Engelhardt had cautioned that beam-to-column connections could crack.

In December 1993, the Journal of Structural Engineering published research by Engelhardt documenting his work conducting seismic lab tests of steel beam and column connections built to current codes. The work, which also involved sifting through the results of lab tests that other researchers had already conducted, stemmed from work he had been hired to do by the Steel Committee of California to find ways to reduce construction costs for beam-column connections.

He didn’t find a cheaper alternative. Instead, he uncovered “a consistent pattern of poor performance” due to quality control issues in welding. And he tried to warn his field and the steel industry.

“We didn’t want to be a Chicken Little saying, ‘The sky is falling.’ But we expressed concern and said, ‘We need to look more closely at this,’” he says.

Sabol says much of the industry initially dismissed Engelhardt’s findings. The Los Angeles Times called it controversial. But when the Northridge earthquake hit one month later, Engelhardt’s theory proved right.

“Engineers knew about research that had been done in the 1960s and ’70s that showed that moment frame connections, the kind we used, were able to undergo a tremendous amount of deformation without fracturing,” Sabol says. “It turns out that our collective memory was a little bit selective. The tests did show excellent performance—if the connection was done in a specific way. If done in a similar but not exact same way, performance was not as good.”

If it hadn’t been for what was observed after Northridge, and the Getty’s commitment to fixing it, nothing might have changed.

“Without Northridge, we probably would have kept building the way we had before—until some other earthquake came along to show us we were wrong,” Engelhardt says.

But there’s still quite a bit of risk unless the city takes stronger action.

Each year, LA conducts a citywide earthquake drill called the ShakeOut. The ShakeOut’s 7.8 hypothetical earthquake is epicentered on the San Andreas Fault, and causes widespread damage across the Southern California region. In this scenario, five steel moment-frame high-rise buildings will collapse and 10 more will be “red-tagged,” meaning they could be deemed too unsafe to enter.

Northridge is the strongest local earthquake that virtually everyone living in Los Angeles has experienced—and it wasn’t that big. “Compared to what could happen in a large earthquake, this was a much more moderate earthquake,” says Sabol. LA hasn’t had a major earthquake in so long that it affects the city’s collective memory when it comes to creating the sense of urgency needed to make changes.

“When we do testing in the lab, it’s not the real thing, and there aren’t enough earthquakes,” Engelhardt says. “That tends to lull us into a sense that everything is going well.”