On 2 July 1937, Amelia Earhart and navigator Fred Noonan left Lae, New Guinea, in a Lockheed Electra 10E, bound for Howland Island. They were on the longest and hardest leg of an attempt to fly around the world. They never arrived. The search that followed was the largest and most expensive in American history to that point, and it found nothing.
What made the leg so difficult
Howland Island is about two kilometres long and barely rises above sea level, roughly 4,000 kilometres from Lae across open Pacific. Finding it required celestial navigation to a precision that left almost no margin, and then a visual sighting.
The US Coast Guard cutter Itasca was stationed at Howland to provide radio bearings. This is where the attempt came apart. Earhart's transmissions were received by the Itasca, but she appeared not to hear the ship's replies, and the radio direction-finding that should have guided her in did not work. A trailing wire antenna that would have improved her low-frequency reception had reportedly been left behind, and there were mismatches between the frequencies each party was using.
Her last confirmed transmissions describe running on a north-south line of position and being low on fuel. Then nothing.
The three theories
Crash and sink. The aircraft ran out of fuel and went into the ocean near Howland. This is the conclusion of the official investigation and the position of most historians. It requires no additional assumptions, and it explains the absence of wreckage — the sea floor there is several kilometres deep.
Nikumaroro. They reached Gardner Island, now Nikumaroro, 600 kilometres to the south, landed on the reef flat and died as castaways. This theory has driven multiple expeditions, and its supporters point to post-loss radio signals, a partial skeleton found in 1940 whose measurements have been reanalysed, and small artefacts. None of it has produced identification, and the artefacts are contested.
Capture by Japan. They came down in Japanese-held territory and were taken prisoner. This has circulated since the 1940s and periodically resurfaces with a photograph or a claimed witness. No documentary evidence in Japanese or American archives has ever supported it, and the specific claims made for it have repeatedly collapsed on examination.
Where it stands
Deep-ocean searches near Howland have been mounted several times with sonar and autonomous vehicles, and periodically something promising appears in the data before failing to confirm. The technology now exists to search the seabed properly. The area is enormous and the target is small.
The honest position is the same as with MH370: the simplest explanation fits the evidence, but until wreckage is identified it is not established. What is established is a chain of small preventable problems — a radio setup that did not work between two parties who each thought it would.