Beginning this fall, bulldozers will cut a single channel across three miles of the Clark Fork and one mile of the Blackfoot floodplains. Natural-looking structures made out of logs, coconut-husk matting, and riprap will keep the river from cutting into the mine waste that remains. Black weed mat will suppress invasive tansy while 150,000 willow, cottonwood and bulrush seedlings, planted by Priest's crews and others, take root. Officials from the Environmental Protection Agency and the state of Montana promise that once the plants have established, the Clark Fork at the Blackfoot confluence will once again be a "real, dynamic river."

Surveying the scene, Priest says it's not that simple, reflecting the disagreement among river restorationists about whether we can return rivers to health by constructing stable, static channels. "Restoration cowboy" Dave Rosgen, the hydrology consultant who prescribed Milltown's meandering channel, says that carving a single, sinuous channel across a river's old path jumpstarts restoration. In the 1980s, in response to 75 years of Army Corps of Engineers "fixing" that straitjacketed thousands of river miles for flood control and navigation, Rosgen came up with a straightforward method for evaluating and redesigning streams. His techniques are now used by agencies from the U.S. Department of Transportation to the Forest Service.

But many geomorphologists, who study how rivers shape the landscape, fault Rosgen's method for using a simplified template instead of a detailed study of how rivers move sediment. They note that river channels evolve in a complex interplay of water flow and the bounce and skid of waterborne sand and gravel. Although the Rosgen method is based on the notion that river channels are stable, aerial surveys show they wind across landscapes unpredictably. Old meanders yield to new meanders, to oxbows, and often intermittently to braids. Season to season, year to year, few rivers stay in the same place. In fact, over-simplifying complex rivers can lead to catastrophe, scientists warn. On Cuneo and Uvas creeks in California, costly Rosgen-inspired reconstruction projects blew out during small floods in 1996 and '97.

The Clark Fork restoration plan recreates a river that never existed, says University of Montana geologist Johnnie Moore. He worries that the reconstructed Clark Fork will be "some kind of reinforced ditch," albeit a curving, natural-looking one. The Clark Fork was once "a true, multi-threaded alluvial river," says Moore. Government surveys dating back to the 1840s depict five distinct channels. Beginning in the 1870s, a mining boom spurred logging, and railroads constricted the channels. Snow melted faster on newly clear-cut slopes; log drives scraped willow-covered islands away, and the river braided even more. Then, in 1908, a massive flood swept across the poisoned wasteland around Anaconda, carrying mountains of tailings that settled in the new Milltown reservoir.

One hundred years later, the river branches across the valley like the veins in a butterfly's wing. So why build a single channel? Money -- and aesthetics. Doug Martin, restoration project manager for the Montana Natural Resources Damage Program, says building a multi-thread channel at Milltown would be trickier, and not cost-effective. What's more, it would mean removing all 6.6 million cubic yards of mine waste. The EPA plans to let ARCO, the company that inherited the Superfund cleanup, leave a third of the waste behind. Besides, a meandering channel fits a cultural ideal of what rivers should look like, says UC Berkeley landscape ecologist Matt Kondolf.