For almost four hours, a camera watched a dense wall of juvenile arapaima moving through an indoor tank. Then the surface broke. Scores of fish rose, exchanged air and dropped back into the shoal before a human observer could comfortably count them.
Researchers recorded 618 such events in a study published in Communications Biology in June 2026. A collective breath lasted 0.83 seconds on average. About 100 fish could participate in an event of that duration, yet the authors often saw more than 100 and on several occasions nearly the entire roughly 200-fish shoal surfaced together.
Those measurements make the synchrony more remarkable, but they also refine the headline. The paper reports an average event duration below one second, not that every event stayed below that threshold. Nor was 100 the largest group observed. It was approximately half the shoal and a useful example of how much activity fitted inside the average burst.
A fish that cannot remain underwater
Arapaima gigas, also known as pirarucu or paiche, is among the largest scaled freshwater fishes. An adult can exceed three meters and 200 kilograms. The animals inhabit the Amazon basin, where warm and seasonally oxygen-poor water helped make air breathing a viable way of life.
An arapaima still has gills, but it is an obligate air breather: the gills do not provide enough oxygen to keep a growing fish alive. A modified, highly vascularized swim bladder functions as its principal air-breathing organ. Research on development has shown how young arapaima shift from water breathing toward this air-dependent system as the swim bladder develops and gill structure changes.
That dependence turns a trip to the surface into a physiological deadline. Oxygen chemoreceptors respond to conditions in the water and blood, while metabolic demand differs from one fish to another. A neighbor can signal that it is time to move, but it cannot take another animal’s breath for it.
The surface also creates danger. A juvenile that breaks cover becomes visible to birds and other predators. Moving with many companions could spread the risk, confuse an attacker or increase the chance that someone notices danger. These are plausible benefits of synchronized breathing, but the experiment did not include predators or measure attack rates.
What “roughly 200” means
The fish were about three months old and close to 30 centimeters long. They had been imported from Peru as fingerlings and raised in a recirculating aquaculture system in Germany. Researchers filmed them through a glass window in their normal rearing tank, which measured three meters long, one meter wide and one meter deep.
The group size was approximate. Facility staff counted the fish weekly, but deaths and possible cannibalism between counts meant the authors could not supply an exact number for the recording day. The water was also too turbid to count every body reliably during a breath.
Instead, observers estimated participation in four bins: less than 20 percent of the shoal, 20 to 50 percent, 51 to 80 percent, or more than 80 percent. In more than 85 percent of the 618 recorded events, participation fell between 20 and 80 percent. The paper also reports frequent events exceeding 100 fish and several involving almost all of them.
The camera ran for 219 minutes on a single day. A collective event began when the first fish released air bubbles while approaching the surface and ended when the last participant dived. The interval to the next event began only after that last fish had gone down.
A hundred breaths inside 0.83 seconds
The mean duration of a whole group event was 0.83 seconds, with a standard deviation of 0.33 seconds. Events involving more fish tended to last slightly longer, but even a breath involving about half the shoal occupied roughly the time an isolated juvenile needed to complete one breath.
Thirty clearly visible fish were sampled from the shoal videos to estimate individual duration within a group event. Their own breathing sequences averaged 0.34 seconds. By comparison, juveniles tested alone took 1.03 seconds on average from releasing old air to diving away.
The group did not merely compress each visit. It also returned to the surface far more often. Collective events were separated by 15.34 seconds on average. When more fish participated, the wait to the next event lengthened slightly, although the correlation was weak.
This is not evidence that 200 fish shared one oxygen requirement. It is evidence that individual requirements were flexible enough to be socially reorganized. Some animals may have surfaced earlier than their physiology demanded, while others may have joined near the point at which waiting longer was costly.
Six isolated fish exposed different clocks
To see what the private rhythm looked like, the team recorded six juveniles separately. Each fish was observed twice for one hour, with a day between sessions. These animals differed consistently in both how often they surfaced and how long they remained in the breathing sequence.
Across the first set of trials, the average interval was 69.6 seconds. In the second, it lengthened to 107.5 seconds, a change the authors interpreted as habituation to the test setting. Individual patterns were irregular rather than metronomic, and the team grouped the six fish broadly as fast, medium and slow breathers.
That isolated sample is small. Six distributions cannot be assumed to capture every physiological type present among roughly 200 fish. The same cohort supplied both observations, but the large-shoal video did not identify each fish or connect an individual’s isolated rhythm to its exact place in a collective event.
Earlier experiments in a social catfish found that group air breathing could be driven by social behavior as well as oxygen demand. In that species, aggression helped determine which fish initiated breaths and how the group followed. The arapaima study extends the broader question to a much larger, strongly shoaling animal, but it did not test aggression as the mechanism.
How a simple rule can create a collective breath
The empirical record showed the pattern. An agent-based model explored how it might arise. The researchers represented each simulated fish as a stochastic, non-periodic oscillator that accumulated motivation to surface according to one of the breathing distributions measured in isolation.
When a simulated fish detected others at the surface, social coupling accelerated its progress toward the breathing threshold. After it breathed, a refractory period made it temporarily unresponsive while it dived and reoriented. This prevented the same burst from immediately triggering another.
Groups made entirely of one breathing type failed to reproduce the observed timing as well as heterogeneous groups. In the simulations, a relatively small minority of fast breathers could pull the collective interval downward. About 10 to 15 percent fast breathers helped produce distributions resembling the empirical record.
The best-fitting version did not force every type into complete synchrony. It allowed stronger response within clusters of fish with similar rhythms and weaker response between clusters. That “cluster synchrony” could preserve some individual constraint while still bringing large subgroups to the surface together.
The researchers have made the simulation data available through Zenodo. A model fit, however, is not a direct observation of the rule inside a fish. The study did not track visual attention, record neural activity or manipulate the proportion of fast breathers in living shoals.
The shoal may be organized in layers
Video showed the juveniles swimming close together in a wall-like or cube-like formation. Fish nearest the surface appeared most likely to take part in the next breath. Lower layers may have been unable to see a surfacing neighbor soon enough, or their internal oxygen state may not have made a response worthwhile.
That spatial picture offers a concrete way for partial participation to emerge. Similar breathers might sort near one another, then react strongly within a local neighborhood. One cluster rises while another remains below and waits for a later event.
It remains an interpretation. The researchers did not reconstruct the three-dimensional position of every fish, and murky water limited visual counts. They identify individual 3D tracking, controlled changes in shoal size and systematic changes in group composition as important next tests.
The proposed predator advantage needs field evidence too. All observations came from one captive shoal in one facility, over one four-hour period, with no aerial attacks. Natural water, vegetation, group density and predator type could all alter whether regular synchronized surfacing is safer than irregular solitary breaths.
What the study establishes
This is the first scientific description of highly synchronized collective air breathing in juvenile arapaima. It documents a repeated event, measures its timing and shows that a captive shoal breathes far more frequently and briefly than six juveniles observed alone.
It does not yet show which fish leads, prove that predation caused the behavior to evolve or demonstrate the proposed cluster mechanism in a natural Amazonian shoal. The approximate group size, binned participation estimates and one-day recording limit the precision of the biological story.
The result is compelling without filling those gaps prematurely. A shoal can coordinate an act whose necessity remains private to every member. What looks like a single breath at the surface is hundreds of separate oxygen budgets finding a narrow moment in common.