The Gallops Island experiments are usually told as a mystery: nobody could transmit influenza, and nobody knows why. That is not quite the shape of 1918. Other people were doing the same experiment in the same months, and some of them were reporting success. The American failure is only strange when you set it beside them, and it is only interpretable when you look at how the subjects differed.
Tunis, September 1918
On 1 September 1918, at the Institut Pasteur in Tunis, Charles Nicolle and Charles Lebailly published a note in the Comptes rendus of the French Academy of Sciences describing what they had done with the sputum of a patient on the third day of influenza.
They passed it through a filter fine enough to hold back bacteria, and gave the filtrate to two healthy volunteers — one by injection under the skin, one into the eyes and nose. Both, they reported, developed influenza. They also gave filtered material to a monkey, which became ill.
Their conclusion was the one that mattered: whatever caused influenza was small enough to pass a filter that stopped bacteria. It was not Pfeiffer’s bacillus. It was what the period called a filterable virus, meaning only “an infectious agent we cannot see or grow.”
Nicolle was not a marginal figure. He had worked out that typhus is carried by the body louse, for which he would be given the Nobel Prize in 1928. When he said a thing passed a filter, people listened.
It is worth being clear about what his experiment actually was: two people. No controls, no comparison group, and the same absence of any laboratory test for influenza that undid the Americans. By modern standards it is not a result at all. But it was the state of the art in September 1918, it was published in the most respectable place available, and it set the agenda for everybody who came after.
Hygienic Laboratory Bulletin No. 123 cites Nicolle and Lebailly twice, in the headnotes to Experiments 5 and 6 of the Boston series, both times with the phrase “following Nicolle and Lebailly.” Those are the two experiments in which filtered secretions and whole blood were injected under the skin of twenty American sailors. The Americans were not groping in the dark. They were repeating a reported success, and it did not repeat.
Japan, December 1918 to March 1919
The other experiment ran at almost exactly the same time as the second Boston series, and it is the one that should be set beside Gallops Island whenever Gallops Island is quoted.
Professor T. Yamanouchi and two colleagues, Dr K. Sakakami and Dr S. Iwashima, published their results in the Lancet of 7 June 1919, under the title “The Infecting Agent in Influenza: An Experimental Research.” It is barely a page long.
They open with the scale of what they were in the middle of. By official statistics, they say, up to the end of January 1919 Japan had seen nearly thirty million cases of influenza, of which 170,000 proved fatal. (The figure is startlingly high for a country of some fifty-six million people, and later historical estimates of the case count are lower; it is given here as they printed it.) Then: “Inasmuch as 52 of our friends, doctors and nurses, offered themselves as subjects for experiment, we have been able to solve some important questions relating to this infectious disease.”
The work ran between 1 December 1918 and the end of March 1919.
Experiments 1 and 2. They made an emulsion in Ringer’s solution of the sputum of 43 influenza patients. Twelve healthy people had it put into the nose and throat. Twelve more had a Berkefeld filtrate of the same emulsion.
The result is given in one paragraph, and every clause of it is load-bearing:
The results of these experiments are very significant; among the subjects treated were six persons who had already had influenza, and all six showed no symptoms of illness. But all of the other 18 subjects, both those who had received the emulsion and those who had received only the filtrate, were attacked by the disease, after an incubation of two or three days. Their fever was sometimes slight (37·5°C.), sometimes very severe (39°C. or more). The subsequent symptoms were headache, sore-throat, lumbago, cough, and the like.
Experiment 3. A filtrate of the blood of influenza patients, into the nose and throat of six more healthy people. “The results were precisely the same as in the previous experiments.”
Experiment 4. Subcutaneous injection: four people with the filtrate of the sputum emulsion, four with the filtrate of patients’ blood. “They all, with the exception of one who had previously had influenza, developed the disease after two or three days’ incubation.”
Experiment 5. A pure culture of Pfeiffer’s bacillus, and a mixture of the bacillus with pneumococci, streptococci, staphylococci, diplococci “and many other like microbes common in the sputa of influenza patients,” into the throat and nose of fourteen healthy people who had never had influenza. “There were no symptoms of illness following these injections.”
Their summary is five numbered lines, and it reaches in the spring of 1919 the position the rest of the world would not settle into until the 1930s: that the agent cannot be removed by filtering; that it infects through the mucous membrane and by inoculation; that it is in the sputum and the blood; that “the known bacilli, such as Pfeiffer’s bacillus, pneumococci, and some diplococci are not the cause of influenza”; and that everyone who has previously had influenza, or has been through their experiment, is immune to it.
Reading the two together
The Japanese paper has exactly the weaknesses of the American one. There was no test for influenza in Tokyo any more than in Boston, so those eighteen cases rest on clinical judgement, made by people who were running an experiment they hoped would succeed, during the largest epidemic in their country’s history, on subjects who were mostly doctors and nurses and therefore exposed to influenza every working day. No chart, no individual data, and no control group of people who were inoculated with nothing. Modern readers should hold it as loosely as they hold Rosenau’s.
And that is the point. Hold both loosely, and you are left with two careful groups, using the same methods, in the same four months, reporting opposite results. Technique does not explain it: both filtered through the same kinds of candle, both pooled sputum from multiple patients, both sprayed noses and injected under skin.
What differed was who was in the room.
The Americans drew their men from a naval camp that had recorded 186 cases of influenza in 1,058 men in the eight weeks before the experiment, and used the previously infected as controls, to show that a dose did nothing to a man who was already immune. Yamanouchi drew his from friends, doctors and nurses, and among them were six who had already had influenza. Those six were the only ones who stayed well.
Set those two sentences next to each other and the mystery of Gallops Island becomes a good deal less mysterious. In the Japanese experiment, prior infection is the variable that explains the outcome: everyone with it was spared, everyone without it fell ill. In the American experiment, prior infection is not a variable at all, because in varying and unmeasurable degree everybody had it.
What became of the question
Nobody settled it with human volunteers. The thing that finally worked was an animal.
In 1931 Richard Shope at Princeton took apart swine influenza and found that pure cultures of the pig’s own influenza bacillus, put up a pig’s nose, did almost nothing: eleven out of thirteen attempts were “entirely negative.” A filtrate of the sick pigs’ material gave a mild illness, and the filtrate and the bacillus together gave the full disease. Shope’s own conclusion was that the two acted in concert, which is not where the science ended up, but he had broken the bacillus’s claim to be the cause on its own.
In 1933 Wilson Smith, Christopher Andrewes and Patrick Laidlaw at Mill Hill in London gave influenza to ferrets using filtered material from human patients, and passed it from ferret to ferret. From that point the agent had a host that could be kept in a cage, infected on purpose, and counted — which is the thing Rosenau never had, and the reason the question could finally be answered.
Nicolle, in Tunis in September 1918, with two volunteers and no controls, had been right about the one thing that mattered. The much larger, much more careful American experiments, with sixty-two men, eight routes and a quarantine island, had been unable to confirm it. In the history of this question, being right and being able to show it are a decade and a half apart.