The fish die first.
They eat zebra and quagga mussels and swim through cladophora, a branching algae that is native to Great Lakes coastlines.
Cladophora has long been a nuisance to humans. In the 1960s and '70s, it bloomed and coughed up thick mats of green weed onto beaches. They rotted, giving off the smell of sewage.
A drop in phosphorus levels brought fewer blooms in the 1980s. But according to the prevailing scientific theory, the introduction of new species — zebra and quagga mussels — helped make the Great Lakes water clearer, multiplying the amount of cladophora by allowing it to receive sunlight in deeper water.
In late summer, when that water is warm, storms break off tendrils of cladophora, which decompose, drawing oxygen from the water. That anaerobic condition triggers the release of the botulism toxin from botulism spores commonly found in fish and aquatic sediments. The mussels filter the toxin from the water and store it. When they are eaten by round gobies - small, blotchy fish native to the Black and Caspian seas — they expose the gobies to a high concentration of the toxin.
Gobies arrived in the ballast water of European ships. The fish quickly thrived, reproducing as many as six times in a summer. By 2002, nearly 10 billion of them were in Lake Erie alone. Gobies poisoned by the botulism toxin become easy prey for migrating waterfowl: gulls, grebes, ducks and loons.
The birds, in turn, are paralyzed. The toxin attacks their voluntary muscle systems, leaving them unable to fly. Ducks in the water drag their wings. In time, their eyes close. Their necks go limp. They drown.
Those deaths are estimated at more than 50,000 in the Great Lakes between 2002 and 2006. The birds die in waves. Nearly 3,000 loons, gulls and red-breasted mergansers littered a several mile stretch of Sleeping Bear Dunes National Lakeshore in 2006. A year later, on Beaver Island, part of Charlevoix County, a smaller die-off washed up four piping plovers — black-necked shorebirds that are protected by the Endangered Species Act.
"That's a heartstring," said Mark Breederland, an extension educator for Michigan Sea Grant, a partnership of Michigan State University and the University of Michigan. "It got people's attention."
The rash of bird deaths has drawn new attention to the possible links between cladophora, quaggas and botulism. Research continues, however, as scientists attempt to verify the theory outlined above.
The anaerobic conditions that foster the botulism toxin are being studied by a coalition of partners, including the state Sea Grant extensions, the U.S. Geological Survey's National Wildlife Health Center, and groups like Common Coast Research and Conservation, a Hancock nonprofit that in 2008 published a field guide for identifying beached waterfowl.
This research has not solved the botulism problem. Scientists are not even sure where the cycle starts, as infected birds can fly significant distances before the paralysis sets in. What they do know is that the end result — more than 25,000 dead birds on the shores of Lake Erie in 2002, and 7,500 more in Lake Michigan in 2007 - poses a far larger ecological problem than weeds stinking up the beaches.
The outbreaks are complicated by the common nature of botulism. The toxin is present in every animal carcass, so biologists cannot automatically rule out other causes of death.
Animals that are infected but not yet dead often are still in, or on, the water. Researchers are not likely to find them. Breederland said identifying the disease in fish is especially complicated, noting, "You need to have a fish that is so fresh its gills are still red."
There is a benefit to the infection's speed. A Type E botulism infection — as opposed to the Type C strain common in Western states — spreads quickly. This limits the risk to humans. "If you catch a fish and it fights on the way in, there's a 99.9 percent chance it's a good fish," said Eric Obert, the extension director for Pennsylvania Sea Grant.
Botulism outbreaks generally end with the waterfowl. There is no evidence of infection in coyotes or other scavenging mammals. Researchers at the U.S. Geological Survey's National Wildlife Health Center know of just one dog that has succumbed to the toxin: a U.S. Fish and Wildlife employee's retriever that gagged on maggots in an infected duck it picked up.
The birds are still a problem.
Scientists have been discussing how best to dispose of them. Some researchers bury them on the beach. Some ship them to funeral homes for cremation. Others send them to the USGS national lab, where biologists inoculate mice with serum or tissue extracts. If the mice die, the bird was carrying the Type E botulism toxin.
The fact that some of those birds washed onto state and national parks is especially troubling. "When this first occurred in 2006, in Sleeping Bear Dunes, it was limited to an 11-mile stretch," Breederland said. "And that's wilderness. It's very pristine.
"We can pick on Green Bay, or on these industrialized parts of Michigan, where you expect some pollution," he said. "But this is a national park. You have to ask yourself, What's going on?" The staff at the USGS lab has asked the same question. But with so many birds spread over so large an area, there is no obvious place to start.
"If we're going to figure out exactly what is going on in the environment, through higher levels of the food chain, including fish-eating birds, the Great Lakes represent a huge area," said David Blehert, a microbiologist at the lab. "We would be better off if we had some idea of where in the Great Lakes we should start looking."
There are patterns to the die-offs. Whitefish Point in Lake Superior, Presque Isle State Park in Lake Erie and the Bruce Peninsula in Lake Huron are all eastern sites relative to the lakes that hold them.
The predominant winds on each lake blow to the east, so it makes sense that the birds wash up there.
What scientists need is an accurate time-and-geography model that can better identify exactly where the birds were infected. Blehert has a plan for that. He wants to float decoy birds on the lakes and track them with GPS transmitters to better understand how a dead bird is carried away from the spot where it is first affected. "We can track the movements of live birds," he said. "To some degree, the movement of a decoy should be easier. You'd put it in at point A and trace it."
The trick is building the right decoy: correctly estimating the speed, buoyancy and appeal to scavengers of an actual dead loon. Blehert might have an answer for that, too. He could just use real dead loons.
 The blooms were caused by high phosphorus levels, brought on by lawn fertilizers and septic systems that drained into the Great Lakes. For more background: http://www.glwi.uwm.edu/research/aquaticecology/cladophora/.
 Type C botulism was first identified in Lake Michigan in 1936. Type E appeared in 1963, according to Michigan Sea Grant. Both varieties are common in spore form in soils and other anaerobic habitats, and in the intestinal tracts of animals.
 The species, Neogobius melanstomus, was first seen in the St. Clair River in 1990.
 Estimate from an underwater camera survey, Johnson, T.B., et. al, Ontario Ministry of Natural Resources.
 Plovers were hunted throughout the 19th century for their feathers, which were used to decorate hats. The Migratory Bird Treaty Act of 1918 protected the species, but environmental changes due to development and beach use continue to depress the population. There are now fewer than 2,000 pairs along the Atlantic coast. For more background: http://www.fws.gov/northeast/pipingplover/pdf/plover.pdf.
 People are more likely to get botulism from food that was not canned properly than from fresh kills. Cooking fish or waterfowl to an internal temperature of 180 degrees should destroy the botulism toxin. Health agencies nonetheless recommend that hunters or anglers not consume animals that appear sick, especially in areas where avian botulism has been reported.
 The process, called a mouse neutralization test, is costly and time-consuming. The USGS lab has partnered with a Wisconsin firm to develop a faster, field-ready test.