On a cold day in June, Scott Pegau leans toward the passenger window of a Cessna floatplane and peers out at the teal waters of Prince William Sound. The glacier-rimmed pocket of seawater on the southern coast of Alaska is protected from the open ocean by a string of rugged islands. It is both moody and alluring. Clouds dally on the snowy peaks and fray against the forested hillsides. The sea is flat and frigid, except for a single row of waves lapping at the rocky shore.

Pegau aims his gaze at the shallow waters behind the breakers. After a few minutes of searching, above a deep bay on one of the outer islands, he finally spots what he’s looking for: a school of juvenile herring. Pegau can distinguish them from other schooling species by the unique way they sparkle – an effect produced by sunlight playing off their silver flanks as the fish bank and roll. Try as I might, I can’t make out any twinkling, just the inky splotch of a few tons of small fish swarming below the surface.

“Small H1,” Pegau says into the headset microphone, tucked snugly under his thick, gray mustache. That’s code for a small school of one-year-old herring. He enters the location on his computer; huddled in the back seat, I make a tick mark on the backup tally. It’s the first of dozens of schools we’ll see on our flight.

Pegau conducts these surveys every year in hopes of understanding what’s in store for the herring population in Prince William Sound. The fish mature and begin to join the spawning stock at the age of three, so the counts give scientists and managers a clue about how many adults may be coming up the pipeline. Researchers and fishers alike always hope the answer will be many. But every year for the past quarter century, they have been disappointed.

The herring population in Prince William Sound crashed in 1993, just four years after the Exxon Valdez oil spill released 11 million gallons of crude into these waters. The collapse put an end to an $8-million-dollar-a-year fishery, and left a hole in the middle of the marine food web. Scientists have spent years trying to understand if and how the spill played a role in the herring’s demise here, and the results have been hotly contested. All of the legal proceedings finally closed in 2015, with herring listed as an impacted species but with most herring fishers feeling poorly compensated.

Even more concerning is the fact that, unlike most species hit by the spill, the herring haven’t bounced back over the decades since. Populations of forage fish are known to boom and bust, so most scientists thought it was only a matter of time before they rebounded. But 25 years later, there’s still no sign of recovery on the horizon.

“There’s definitely a possibility that the ecosystem went through a tipping point,” says Pegau, who coordinates the herring program at the Prince William Sound Science Center, an independent research institute whose work is funded in part by money from the spill settlement. A host of factors, which scientists are still trying to untangle, could be to blame, from hungry whales to virulent disease. “There’s no one thing that’s keeping them down,” Pegau says. “I think pretty much everyone is convinced of that.”

The herring mystery is a maddeningly concrete example of the often-abstract interconnectedness of nature, which frequently gives ecosystems their resilience, but can sometimes make them rebellious. If jolted a particular way – for instance, by a human-caused disaster or an environmental shift – an ecosystem may not revert to its original state. Instead it may settle into a new normal, leaving both the organisms and economies that rely on it reeling.

But the herring in Prince William Sound may also hold clues to long-standing questions about why ecosystems change, and how they cause fish populations to flourish or founder. After 25 years of research, scientists have collected reams of data on the herring, and half a dozen hypotheses to explain their plight. The data have yet to reveal satisfying answers, but what researchers learn about ecological resilience – and the true value of a species – could have ramifications well beyond Alaska’s shores.

After a few hours of surveying, the floatplane splashes down in Eyak Lake, which fills the Y-shaped valley separating Cordova from the looming peaks of the Chugach Mountains. “We survived another one,” jokes the pilot, as we taxi to a floating dock and unload. Pegau makes plans to fly again the next day, and we climb into the well-worn seats of his white F-250 for the short drive back to town.

Cordova is a no-frills community on the far eastern edge of Prince William Sound that has weathered its share of ups and downs. It was established in 1906 as the seaport from which to ship vast quantities of copper mined in Kennecott, 200 miles (321 km) inland. When the mine closed in 1938, Cordova fell back on its reputation as the self-proclaimed razor clam capital of the world. But that industry went belly up on March 27, 1964 – Good Friday – after a magnitude 9.2 earthquake heaved the clam beds 6 feet (2 m) above sea level.

That left fishing and canning as the main games in town. Both revolved around salmon and herring. While the much smaller herring rarely exceed 10 inches (25 cm) in length, they are prized for their roe, as well as for food and bait.

But in 1989, again on Good Friday, the Exxon Valdez oil spill cast the town’s future into question once again. Many residents considered it the worst blow yet. “There’s a natural disaster, and there’s a manmade disaster,” says Sylvia Lange, a Cordova native and longtime fisher who now runs a local hotel. She has experienced both kinds of catastrophes in her life, and the two, she says, feel completely different.

The ecological impacts of the spill, by now, are legend. Despite the efforts of thousands of response workers, wildlife including sea birds and otters perished in droves. The financial impacts followed swiftly, too. The cleanup effort provided some business, but tourism lagged and managers closed fisheries until they knew how the spill had affected fish populations. Meanwhile, for unrelated reasons, the price of pink salmon tanked. “Cordova was in a deep, deep depression,” Lange says, “not only psychologically but economically.”

At first, it seemed like the herring made it out unscathed. The Alaska Department of Fish and Game reopened the fishery in 1990, and in the two years that followed, herring fishers had some of their best seasons on record. But when the spring of 1993 rolled around, the herring all but disappeared. The population dropped from more than 100,000 metric tons to less than 30,000 tons.

The crash devastated a community still coping with disaster. John Renner, a longtime Cordova fisher and chairman of the herring division at Cordova District Fishermen United, says the spring herring roe harvest was a staple of Cordova’s economy. “The whole town depended on it for the first shot of revenue of the season,” Renner says “People paid their taxes, got out of debt, that type of stuff.” For many, it was a big part of their livelihood; Renner once made a quarter of his annual income off of herring roe, and just talking about the collapse still makes him angry. A draft report commissioned by the science center estimates that, in total, losing the herring has cost Cordova almost $200 million and the region nearly $1 billion.

The passage of time has done little to settle questions about what caused the herring crash. Initially, many scientists doubted whether the oil spill could have caused a decline four years later. Some early studies also suggested the impact of the oil was minimal, including those by Walt Pearson, then a fisheries biologist with Battelle Marine Sciences Laboratory whose work was funded by Exxon. Pearson’s research found that adult herring could only have been exposed to low levels of oil for a short window of time, and that there had been little overlap between oiled beaches and herring spawning grounds. “The effects were quite localized,” Pearson says. He concluded that the biggest factor contributing to the crash was that there were too many herring and not enough food, due to a natural shift in ocean conditions.

Many fishers didn’t buy that. “Anyone with half a brain would figure out it was oil,” says Jerry McCune, president of Cordova District Fishermen United, a union-turned-nonprofit advocacy organization. The herring spawn occurred just weeks after the spill, and McCune, Renner, and others think that the oil devastated the cohort of herring born in 1989. They say the failure of those fish to show up in 1993 accounted for the collapse.

But survey data collected during the crash suggest that it affected fish of all ages, says Pegau, not just the young ones. And a recent statistical analysis found little evidence for any direct impacts of the spill. Instead, Pegau and others think that if the oil did play a role in the collapse, it probably did so by weakening the herring, or the food sources upon which they depend, making it easier for something else to do them in.

The prime suspect, in Pegau’s estimation, is a disease called viral hemorrhagic septicemia (VHS). While there was no official monitoring program then, fishers and scientists saw signs of VHS in 1993. “It can take a population down in a real big hurry,” Pegau says. As the name implies, fish with VHS hemorrhage and can die from organ failure. The disease spreads quickly through herring’s dense schools or when fishers corral them into an enclosure to harvest their spawn, as local fishers were doing around the time of the spill. Some researchers think that this practice, called pounding, combined with the high herring numbers before the crash, contributed to a deadly outbreak of VHS.

But the risk of an outbreak could have been exacerbated by the spill, too. Fish embryos that don’t die when exposed to oil may carry genetic scars, particularly in something called the aryl hydrocarbon receptor gene. “It turns out that that gene gets completely knocked out among survivors,” says Paul Hershberger, a disease ecologist with the U.S. Geological Survey. And compromising that gene may weaken the immune system in fish, potentially making them more susceptible to disease. Hershberger’s colleagues have demonstrated this effect in Atlantic killifish, and now, his team is testing it in herring.

Exposing herring embryos to oil may also cause them to develop heart defects that put them at a general disadvantage. They can’t swim as fast or as long, which makes them more likely to get eaten, says John Incardona, a toxicologist at NOAA’s Northwest Fisheries Science Center and lead author of a study on this subject published in 2015. In lab experiments, Incardona has found that developmental effects occur even when herring embryos encounter levels of pollution far below what is generally considered harmful. “We think all of us are way underestimating what the initial injury was to herring,” he says.

Richard Thorne, a fisheries scientist who recently retired from the Prince William Sound Science Center, takes issue with the idea of a delayed collapse altogether. Evidence that the herring population remained high until the 1993 crash came from the Alaska Department of Fish and Game’s population estimates, which are based on the stock assessment models the state uses to manage its fisheries responsibly and set sustainable harvests. But in 1993, Thorne started conducting acoustic surveys of the herring population, and realized his numbers lined up best with a different set of data collected by Fish and Game: observations of how many miles of shoreline were covered in herring spawn. Looking back at this pre-crash spawning record, Thorne came up with an alternative population history, which suggested that herring numbers started falling immediately after the spill. He thinks the fish died from ingesting oil and that the collapse, if there was one, resulted from allowing fishers to harvest a herring stock in the early ’90s that managers didn’t yet realize was already declining.

Pegau, for one, doesn’t think scientists will ever know what actually transpired. “We’ll never be able to say one way or the other because no one was collecting data when it happened,” Pegau says. And frankly, he doesn’t really care what caused the initial collapse. The more pressing question, Pegau says, is why the herring haven’t come back.

I meet John Platt on a floating dock in the old harbor, and the first thing he says as he shakes my hand is, “Why are we still talking about herring 25 years later?” Platt is a third-generation Cordova fisher with a leathery face and the gnarled physique of a former wrestler. And he’s being coy; we both know the answer to his question.

Platt used to fish for herring, collecting them using a type of net called a purse seine to harvest roe. He gamely drives me out to see the net, which he stores 10 minutes outside of town and which is – as far as I can tell – the only piece of herring gear left in all of Cordova. “I always thought they would come back,” he says. We pull up to a rusted white pickup truck overgrown by weeds and Platt gestures toward the sorry sight before us. “This basically sums up the herring fishery.” He gets out and starts to unwrap a battered blue tarp covering a lumpy mass on the truck’s flatbed, He finally tugs free a loop of black mesh for me to see. The net – which cost about $20,000 new – still looks good decades later. After all, it’s hardly been used.

Like many, Platt was nearly destroyed by the herring collapse. Commercial fishing permits in Alaska are traded like stocks; the state issues a limited number, and fishers buy and sell them at prices that generally reflect the value of the fishery. And the seine fishery was a high-stakes gamble. Herring roe was a hot commodity and fishers like Platt jockeyed for position near schools of herring so they could scoop them up when the fishery opened, sometimes only for an adrenaline-filled 15-minute window.

Before the crash, when Platt bought his seine permit, the going price was nearly a quarter of a million dollars. Today, the same permit is worth just $31,000.

Some note that’s remarkably high, given that the roe market has deflated and there have only been two modest herring harvests in Prince William Sound since the crash, in 1997 and 1998, when managers thought the fish might be making a comeback. But the permit’s unsinkable value is of little use to people like Platt, who purchased his permit with a loan from the state, and struggled to make the payments – and pay taxes – without any fish to catch. In total, those who held permits for herring in Prince William Sound took a $50 million to $60 million-dollar hit in lost permit values, according to a recent economic analysis.

Platt got paid by Exxon for working on the cleanup effort immediately after the spill, but that money went straight into paying off boats and gear. To settle his permit debt, Platt ultimately had to sell off his salmon boat and turn over the payout from a class-action lawsuit against the company, which, after going before the Supreme Court in 2008, came in at a fraction of the original award. For many, though, it was too little, too late; the loss of the herring had already taken its toll. “It caused divorces, ruin, a few people killed themselves,” Platt says.

Like many, Platt thought the failure of the herring to recover might be grounds for reopening the 1991 settlement between the government and Exxon (now ExxonMobil), which closed before the herring crashed. The settlement included $900 million in payments, in addition to criminal fines, and a clause that would make an additional $100 million available for long-term impacts that weren’t considered in the original agreement. “This is textbook, what it was for,” Platt says.

In 2006, government lawyers did launch an effort to file a claim under the reopener, but it was later aborted. Moreover, the claim made no mention of herring. “It’s maddening,” Platt says. But Pegau thinks there was a simple reason: Linking the fish’s poor recovery to the spill would have been a hard case to make.

Traces of oil still remain in Prince William Sound, buried a few feet in the sediment among beach pebbles and sand, but most scientists say it has little ecological impact on herring today. Indeed, if the spill had any role in the fish’s demise, it was by helping to knock the population off a cliff in the first place. Other forces have now taken over and seem to be holding the herring down. And they don’t seem to be letting up. In 2015, after what seemed like a few promising years, the herring population dropped again to around 8,800 U.S. tons (8,000 metric tons) – less than half of what it was after the crash in 1993.

“I think the system reset itself,” says Ron Heintz, a nutritional ecologist in Juneau with NOAA’s Alaska Fisheries Science Center. “We ended up in a new state that apparently doesn’t include herring.”

One factor at play is predation. Forage fish, by definition, get eaten, and herring are no exception. “It’s a critical food resource,” says Mary Anne Bishop, an ecologist at the Prince William Sound Science Center. She describes the spring spawn as a frenzied feast when the herring turn coastal waters white with eggs – each female releases about 20,000 of them each year. “It’s the whales, it’s the sea lions, the harbor seals, all the birds coming in,” she says. Predation doesn’t stop as herring age, either; dozens of species consume them throughout their life cycle.

While there may have been enough herring to fill bellies and nets when the fish were plentiful, they may now be trapped in what scientists call a “predator pit.” After everyone has had their fill, there simply aren’t enough fish left for the herring population to climb out of the hole. More young adults join the spawning stock each year, but not enough to outweigh the number being eaten.

There’s debate about which animals are doing the most damage, but humpback whales are a possible culprit. Their numbers have quintupled in Prince William Sound in recent decades as the gentle giants have recovered from whaling. Scientists say the whales here have learned to specialize in herring, sometimes banding together to trap the fish in “bubble nets” before taking turns gulping them down en masse. Studies suggest humpbacks may consume 20 to 75 percent of the spawning herring population each year – the equivalent of the fishers’ historic share and then some.

Other scientists, including Pearson, have suggested that salmon hatcheries may bear much of the blame. Starting in the late 1970s, managers began releasing hatchery-raised pink salmon into the sound, and in the 1980s, they began to ramp up the numbers. Researchers have hypothesized that young salmon may eat or compete with juvenile herring for food, while older salmon returning from the sea may eat herring of all ages.

It also appears that the herring in Prince William Sound continue to see diseases like VHS more than their neighbors. Hershberger developed a test to detect whether fish have recently encountered disease, and found consistently higher levels of exposure in Prince William Sound than in Sitka, 450 miles to the southeast. Some wonder whether, as Platt puts it, the herring here are “wimpy” because of some lingering epigenetic effect of the spill that has been passed down from generation to generation. Hershberger says no one has tested that yet. “At this point, all we can do is speculate.”

Factors like predation, competition, and disease can limit populations from the top down. But that’s only half of the story: There has also been a low supply of young herring. Researchers call this poor recruitment, and they suspect it’s the result of environmental factors limiting the population from the bottom up.

For the herring to recover, they need a few big years in close succession to overwhelm the demands of predation and escape the threat of disease. But that hasn’t happened in a long time, Pegau says. “It’s been 25 years of bad luck.” He points to sweeping natural changes in the North Pacific in 1989 – the same year as the spill – as a potential turning point. The exact nature of the shift was complicated, with some parts of the ocean warming and others cooling, but the impact on marine organisms was pronounced. Across the Gulf of Alaska, animals toward the bottom of the food web, like shrimp, crab and herring have fared poorly since, while larger fish like halibut and cod have multiplied. This apparent contradiction continues to stump Pegau. “I have yet to figure out what in the world supports them,” he says of the thriving predators.

Much of the science center’s research looks at how oceanographic conditions affect herring in hopes of understanding why the tide has turned against them in recent decades. Among other things, that involves tracking where currents carry larvae, determining which environments young fish inhabit, and studying what controls the quantity and quality of food available to them as they store up energy to survive the harsh winter.

New research also suggests that herring recruitment may be linked to the amount of freshwater that pours into the Gulf of Alaska. High discharge years correlate with recruitment failures, says Eric Ward, a statistician at the Northwest Fisheries Science Center who led the study, published earlier this year. The mechanism, though still unclear, may have to do with how freshwater from rainfall and melting ice affects the strength and timing of the spring plankton bloom – the flurry of photosynthesis that kickstarts the entire ecosystem every year. In recent decades, there have been fewer years with extremely low runoff, which correlate with upticks in herring recruitment – what Ward calls “herring baby booms.” And as climate change melts glaciers and messes with rainfall patterns, the trend may continue.

These oceanographic factors may help explain why herring recruitment has also been weak in other parts of Alaska since the early 1990s. The reason places like Sitka still have a healthy herring fishery despite these changes, Pegau says, may simply be that the population there never crashed in the first place.

Scientists hope that figuring out what’s going on with herring will shed light on bigger questions about what fish need for successful recruitment – a problem that has stumped researchers for decades. Trevor Branch, a fisheries scientist at the University of Washington who studies the herring, says it’s possible that a whole host of things have to line up for successful recruitment: the right water temperatures, the right salinity, and abundant food, among them.

If scientists are to have any shot at figuring out how these fit together, they need lots of data collected over many years. And the research sparked by the Exxon Valdez oil spill and the subsequent herring crash have furnished just that. “If ever we were able to pinpoint something, it would be with Prince William Sound herring,” Branch says.

The day of my second flight with Pegau, the weather is sublime. We head across Prince William Sound to survey its far southwestern corner. A cruise ship glides beneath us, then a cluster of salmon seiners. The small skiffs that accompany them trace lazy circles on the surface, like ripples from giant raindrops, as they loop their nets around unsuspecting fish.

We officially begin the survey in a milky, ice-flecked fjord with a reclusive glacier tucked away at its head, and follow the coastline in and out of emerald bays. In the narrow passages between Evans, Elrington, and Latouche Islands, we spot school after school of herring, clustered along the rugged shoreline. In the afternoon light, I finally see them sparkle.

Nearly 100 years ago, these were prime fishing grounds for an earlier incarnation of the herring fishery. Fishers caught huge quantities of herring, which were reduced for oil. For five consecutive years, they brought in in an average of 40,000 tons a year, Pegau says, marveling at the scale. Catch records show that those big hauls likely drove the herring to collapse, but remarkably, they appear to have recovered in the span of just three or four years. Pegau takes this as evidence that the herring in Prince William Sound have rallied back from the brink before, like others around the world.

John Trochta, one of Trevor Branch’s graduate students at the University of Washington, has analyzed more than 50 historical herring populations across the globe, most of which have collapsed at some point. He found that the majority rebounded within a decade, but there were a few exceptions where herring numbers remained low for at least twenty years after a crash. One is in Prince William Sound; another is off the coast of Japan and southeastern Russia. There, fishers once harvested nearly a million metric tons of fish per year from the legendary Hokkaido-Sakhalin stock. But by the 1930s, perhaps due to intense fishing pressure and oceanographic changes, the stock began to decline sharply, until, by 1955, there were hardly any fish left. It’s the only herring stock yet to come back after more than 60 years. No one knows whether the same fate awaits the herring in Prince William Sound.

On my final day in Cordova, I stop by Pegau’s office at the science center – a converted icehouse perched on stilts just inside the entrance to Cordova’s harbor. From Pegau’s second-story window, he has a clear view of the bustling docks and the mountains that stand guard over town. With this year’s survey nearly complete, I ask him how it looks. “This is probably the best year we’ve seen,” he admits. He saw a lot of schools, and the schools held a lot of fish. Still, he’s reluctant to wager whether that bodes well for herring. “I’ve felt optimistic in the past; now I’m a lot more reserved.”

There’s nothing Pegau can do to help the herring. There’s no fishery to manage, or acute environmental stress to relieve. That’s not the way he sees his job anyway. His goal is to understand the vulnerability and value of the fish so that other scientists and managers around the world can be better equipped to do the same. Increasingly, he and others think that the answer lies in studying an ecosystem as a whole, and how an individual species like the herring fits in. And in that regard, he is more hopeful than ever. “It’s a great puzzle,” Pegau says, his dark eyes twinkling with excitement. “One of the real joys is to see how all those pieces fit together.”

This story originally appeared in bioGraphic, an online magazine about nature and sustainability powered by the California Academy of Sciences.