Dr. Frederick Russell, director of the International Health Division, believed that the yellow fever viruses from Africa and South America should be brought together at a central laboratory--away from the field labs of Nigeria and Brazil--and tested for cross-immunity. He chose Sawyer to head this new venture. Several rooms at the Rockefeller Institute were provided for the lab space, and in June 1928, Sawyer and two researchers, Drs. Wray Lloyd and Stuart Kitchen, began the dangerous work.

Sawyer's research team collected three strains of yellow fever virus, the "Asibi," the "French" (which Harvard bacteriologist Andrew Sellards had obtained from a patient at the Institut Pasteur in Dakar), and the "Brazilian". Preserving the live virus in blood or tissue during the long journey to the United States was a challenge. Sawyer discovered that extended preservation was possible if the infected monkey blood was freeze-dried and stored in sealed glass tubes. During the first year Sawyer and his staff proved through cross-immunity tests that all three virus strains produced protective antibodies in rhesus monkeys; if not identical viruses, they were very similar. Further tests showed that the disease caused by Noguchi's Leptospira was not yellow fever, but Weil's disease, which also causes jaundice.

They then turned their attention to developing a vaccine. This was badly needed; by early 1929, ten researchers in various Berlin and London labs had reportedly contracted yellow fever from working with infected monkey blood and tissue. In April that year, Sawyer himself caught the disease, despite elaborate precautions. It was not a severe case, but he was ill for several weeks, and it was five weeks before he returned to the lab. Many of the lab staff contracted it over the following two years, though none died.

Meanwhile, Max Theiler, working in Sellards's Harvard lab, had begun developing an attenuated yellow fever virus using mouse brains. Mice, it turned out, did not develop classic yellow fever symptoms when injected abdominally, but did develop encephalitis (inflammation of the brain) if the virus was injected directly into the brain. When the virus was passed from mouse to mouse (by transferring infected brain tissue to successive mice), it produced increasingly severe encephalitis, but when each "pass" was injected into rhesus monkeys, the virus was progressively less virulent to the liver and other organs. That is, the power of the virus had been weakened or attenuated. To prove that the mice had yellow fever and not some other encephalitis, Theiler demonstrated that if the virus were mixed with serum from a yellow fever-immune person and injected into the mice, it would protect them from the encephalitis. If non-immune serum was used, the mice developed encephalitis. When Theiler published these findings in 1930, Sawyer and Russell visited him to learn more. Soon afterwards, Theiler joined the team at the RF Yellow Fever Laboratory.

Within several months, Sawyer and his colleagues had developed a vaccine using the attenuated virus. The vaccine had two parts: a ten-percent suspension of mouse-brain tissue with yellow fever virus in fresh sterile human serum, and human immune serum from people recently recovered from yellow fever (i.e., many of the lab staff, including Sawyer and Theiler). These were injected simultaneously. Dr. D. Bruce Wilson, recently back from Brazil, volunteered to be the first human test case. With no ill effects beyond soreness at the injection sites, Wilson had a good level of immunity within several days. Because the vaccine required human immune serum, it was not suitable for large-scale production. But the RF was able to vaccinate eighty-five lab workers and field staff during the next four years, and put an end the accidental and sometimes fatal infections.

Sawyer and his colleagues also were able to modify the test Theiler had used to verify that yellow fever caused the encephalitis in mice, and to use it as a tool for mapping where yellow fever had been. Their "mouse protection test" mixed serum from people in reported yellow fever areas with the lab virus and injected it into the brain of a lab mouse. If the mouse did not develop encephalitis, it indicated that the serum donor had had yellow fever at some point; the donor's yellow fever antibodies had protected the mouse. Between 1931 and 1936, Sawyer supervised a worldwide survey of yellow fever, which provided public health workers with a much more accurate picture of yellow fever epidemiology during the previous several generations.

In the meantime, Theiler continued to work on further attenuating the virus, hoping to eliminate its ability to produce encephalitis, thus making it safe enough to use for mass vaccination. He and his colleagues also searched for ways to grow the virus in something simpler than live mouse brains. (Unlike bacteria, which can be grown in non-living material, viruses need to be in some sort of living tissue to survive and multiply.) In 1932 Theiler and Dr. Eugen Haagen succeeded in propagating yellow fever virus in several living tissue cultures, the best of which was chick-embryo cultures.

From 1934 to 1936, the team cultured several strains of virus using several different living animal tissues, periodically checking their effects on rhesus monkeys. By late 1936, they had produced an attenuated yellow fever virus, which they called 17D. It caused no encephalitis or yellow fever symptoms, yet stimulated antibody production very well.

The lab immediately began preparing to produce vaccine with the 17D strain. Sawyer had become director of the IHD in 1935, leaving Johannes Bauer in charge of the yellow fever lab. He kept in close contact with the research team, however, and as head of the division, made the ultimate decisions about vaccine production and field trials. The first field trials were done in Brazil, starting early in 1937. By mid-1940 nearly two million Brazilians had been immunized.