Fallingwater, Frank Lloyd Wright's masterpiece, has been featured on a U.S. postage stamp, on the cover of Time magazine, and in perhaps a shelf-ful of coffee table books. Since the house was opened to the public in 1964, some 3.5 million visitors have made the pilgrimage to its sylvan location off Route 381 in the hills of western Pennsylvania. Wright's stunning creation-the living room with its mound of native sandstone swelling up through the floor, the ribbons of red-trimmed steel windows, and, above all, those daring cantilevered planes hovering over the falls of Bear Run-has enchanted architectural critics and the general public since its completion 64 years ago.

Too bad it wasn't built properly. While millions of people have a nodding acquaintance with Fallingwater, relatively few of them know the latest, most dramatic chapter in its history: Only an ingenious repair, completed last spring, removed the growing risk that a large chunk of the house might tumble into the creek below.

The problem had been brewing since the beginning. The concrete beams that support the hung-in-space living room, its two adjoining terraces, and the master bedroom terrace above were too weak for the load they needed to carry. The beams sagged from the moment their supports were removed during construction and continued to droop a little more each year, until by 1994 they were an alarming 4 to 7 inches out of level. It took five years of assessment and planning, four months of actual work, and $1.5 million to stabilize the structure, but now, at last, Fallingwater is secure in its airy perch.

When Wright wrote in 1955 that "Fallingwater is a great blessing-one of the great blessings to be experienced here on earth," it wasn't just more hyperbole from the legendary self-promoter. "The house has a real presence in the American psyche," says Richard Cleary, an architectural historian and Wright scholar at the University of Texas at Austin. "That view of it over the waterfall is a sort of perfect American fantasy. It showed we could have both technology and the natural world, and make them work beautifully together."

A tour of the house reveals Wright's brilliant integration of building and site, so seamless and confident that the dicey engineering seems somehow forgivable. The structure's three levels echo the rock ledges over which the falls flow, as does the rough stonework inside and out. The flagstone floors, always waxed to a high gloss, evoke the wet river rocks. Even the low ceilings-in some places just 6 feet 4 inches high-direct attention outward, to the maples and rhododendrons clinging to the steep hillsides. Good ideas have flowed out of this house and into the American building vocabulary for decades. From multipurpose great rooms to window walls to the foam rubber (a novelty in the 1930s) that cushions its custom-built furniture, there's a bit of Fallingwater in virtually every American home built since.

Yet in 1934, when the project started, Wright was an unlikely choice of architect. At the time Edgar J. Kaufmann, a wealth Pittsburgh department store owner, and his wife, Liliane, decided to build a weekend house, the Depression had trimmed the ranks of clients who could afford Wright's radical, expensive designs. In fact, many believed the 67-year-old was in the twilight of his career. It was the Kaufmann's son, Edgar junior, then working as an apprentice at Taliesin, Wright's Wisconsin studio, who persuaded his parents that they needed the master's bold, mature aesthetic. In December 1934, Wright traveled to the stream-side site the Kaufmanns had selected and was instantly charmed, "The visit to the waterfall in the woods stays with me, and a domicile has taken vague shape in my mind to the music of the stream" he wrote to them a few weeks later.

But it wasn't until late summer 1935 that Wright, prompted by his client's imminent arrival at his studio finally put his vision on paper. Those first sketches, created in a matter of hours, depicted Fallingwater almost exactly as it would be built. While Kaufmann had expected the house to be situated below and to the side of the creek, Wright perched it on the boulders above the falls, leaning out nearly 15 feet over the rushing water. The illusion was perfect: The stream would appear to have eroded Fallingwater's foundation over centuries, lending the house the instant gravitas of a geologically ancient outcrop.

While the idea for Fallingwater was inspired, the execution was fraught with difficulties. The remote site proved inaccessible to concrete trucks, forcing workers to make thousands of small batches by hand, with shovels and rum mixers. A quarry was opened on-site to extract native sandstone or the masonry walls. Skilled labor was scarce, as the just-formed Works Progress Administration siphoned off the region's best stonemasons for government jobs.

But the biggest problem was the 15-foot cantilevers. Made of concrete, the beams projecting over the creek needed additional reinforcement just to support their own weight, not to mention the loads pressing down on them form above. Wright ordered a slab poured under the beams to stiffen them from below and steel reinforcing rods (rebar) embedded primarily along the tops of the beams to prevent them from cracking.

Suspicions soon arose at the building site that Wright's engineer, Mendel Glickman, had badly bungled a crucial aspect of the design. The plans called for just with 1-inch-thick steel rods within each of the four main concrete cantilever beams. Metzger-Richardson, the Pittsburgh engineering firm that supplied the steel, urged Kaufmann and the general contractor, Walter J. Hall, to double that number to 16. Hall complied, with Kaufmann's approval, a move that infuriated Wright, who thought that the extra steel would only add to the load on the beams. He wrote to Kaufmann: "I have put so much more into this house than you or any client has a right to expect that if I haven't your confidence-to hell with the whole thing." Kaufmann smoothed Wright's ruffled feathers, but the rods remained. A good thing, too. "Without that extra steel, the cantilever definitely would have failed," says Robert Silman, president of Robert Silman Associates, the structural engineering firm that devised the means to stabilize the cantilevers and save the building from collapse.

Even with the added reinforcement, the living room dipped nearly 2 inches when workers pulled away the cantilever's temporary timber supports in 1936. Ominous cracks quickly opened in the parapet walls of the master bedroom terrace directly above. After feverish analysis, Metzger-Richardson argued for permanent steel bracing to be placed in the streambed below. Wright stubbornly defended his original vision. "I have assured you, time and again, that the structure is sounds," he wrote to Kaufmann in January 1937.

In the end, Kaufmann sided with the architect. The first floor remained brace-free, and construction moved ahead even as costs spiraled. The original budge of $40,000 had seemed generous in a time when a four-bedroom brick home could be built for about $4,000. But the difficult site, endless wrangling between a determined Wright and an equally willful Kaufmann, and the dearth of skilled labor pushed the final cost to $155,000, equivalent to nearly $2 million today.

Fortunately, the Kaufmanns could afford it. And they adored the finished house, which quickly became a magnet for the elite of the time. But the ever-drooping terraces and widening cracks worried Edgar senior until his death in 1955. Edgar junior deeded the house to the Western Pennsylvania Conservancy in 1963, and by 1994 its directors had grown concerned enough about the house's problems to hire Silman's engineers to get a clearer picture. Using impulse radar, ultrasound, and high-resolution magnetic detection, they were able to gauge the condition of the cantilever beams. The findings were sobering: "Stresses were at critical levels. The steel in the concrete beams was still stretching," Silman says. "The building was at the point of incipient failure."

How Glickman could have made such a mistake puzzles Silman. "Mendel Glickman was a first-rate engineer," he says, "and this was a calculation any first-year structural engineering student could do." Silman theorizes that Glickman and Wright underestimated the load of the master bedroom terrace on the cantilever below. During construction, four fat steel mullions were inserted between the windows in the living room to help hold up the terrace, but these vertical supports also transferred the terrace's massive 50-ton load to the outer ends of the cantilever beams, leaving them dangerously overstressed.

Today, assigning blame is tricky. The house's challenging design, combined with slow prewar communications (plans and instruction usually traveled from Wright's Wisconsin studio to the job site via overland mail), made mistakes almost inevitable. "Wright was definitely pushing the envelope with his building, so it's hard to know exactly what happened," concludes Fallingwater museum programs assistant Clinton Piper.

The fact is many of Wright's creations simply outran the technology available at the time. (Another example can be seen in the reconstruction of his dramatic but chronically leaky roof at Wingspread, near Racine, Wisconsin. While few doubt the genius of Wright's designs, it is perhaps a blessing that many of his most ambitious concepts, including whole cities and a mile-high office tower, were never built.

Stabilizing the cantilevers wasn't the end of the work needed at Fallingwater. Leaking roofs had to be peeled back and waterproofed, parapet walls reinforced, and every steel window frame stripped and repainted. Repairs and upgrades to the house and property, including building an on-site sewage treatment plant and extensive landscaping, totaled $11.5 million, funded through private, corporate, and government donations. Still, no on involved in the restoration effort holds any grudge against Wright. "It's a very complex house, and to repair it in a respectful way has been challenging," says Lynda Waggoner, director of Fallingwater. "Now, it looks as good as it did 64 years ago."

Stopping the Sag Cables stabilize Fallingwater's famous-and famously drooping-cantilevers

For more than 60 years, four massive concrete cantilever beams projecting 15 feet into thin air held up Fallingwater's cantilevered terraces and living room. But the unrelenting loads were so great that, over time, the cantilevers deflected, cracking the beams and causing the steel reinforcing bars embedded in the concrete to stretch. Structural engineers Robert Silman Associates examined the building in 1994 and concluded that unless action was taken, the rebar would continue stretching to the point where the cantilevers might collapse. Silman's solution was "post-tensioning," a technique that has been used to shore up sagging concrete bridges. In essence, the process involves securing a length of steel cable to both ends of a section of concrete, then pulling the cable taut. The cable acts like a giant clamp, compressing the concrete so it resists further bending. Prepping the floor Before repairs even began, temporary steel bracing was placed in the streambed to support the ends of the beams. Then, in November 2001, workers carefully numbered and removed each of the 600 waxed flagstones on the living room floor and pulled up the redwood subfloor, revealing the underlying gridwork of concrete beams and joists. Next, the post-tensioning contractor, VSL, threaded a bundle of thirteen ½-inch steel cables alongside each drooping cantilever beam. One end of each bundle was secured in a concrete anchor block attached to both the beam and the house's foundation pier. At the other, unsupported end of the beam, the cables passed through another anchor block and exited through the terrace wall. Near the middle, the cables passed over a concrete deviator block, anchored to the beam above the outermost end of the pier. In theory at least, pulling on the exposed ends of the cables with sufficient force would raise the beam and transfer its stresses back to the foundation. Things get tense Over a three-day period last March, the VSL crew gingerly put theory into practice. Standing on scaffolding set above the creek, they attached powerful hydraulic jacks to the exposed cable ends and slowly pulled. "It was hard to say who was under more stress: the beams, the contractor, or the engineers," says Silman associate John Matteo, who supervised the operation. But the old concrete held, and the beam ends rose slightly, just as planned. Each of the now ultrataut cable strands was then fixed to a metal cone and grouted into the anchor blocks on the beam ends. VSL used this same technique to stabilize some of the concrete joists, running perpendicular to the beams, that support the living room terraces. The post-tensioning did not raise the cantilevers back up to level, nor was the Silman's aim. "We brought them up only about ¾ inch," he says. Raising them any higher would have opened new cracks in the long-settled structure. Besides, Silman says, "That sag is part of the story of the building.

A hydralic jack, attached where the tensioning cables exited the building, gradually tightened the cabels until they exerted a static pull of 195 tons on each side of the beam, lifting it about 3/4 inch. Then the cable ends wer embedded in the anchor blocks and trimmed, and the holes were patched.