A bearing of approximately 201 degrees was obtained on a signal at 3105 kHz described⁶⁴ by the Midway supervisor as

...a faint wobbly fone ... of such a short duration that it was impossible to narrow it down properly.

Table 8: Signal Data for Bearing 4

Source	Sun Set Z	Sun Rise Z	Source Brg	Recip Brg	Midway Sunset: 0650Z Sunrise: 1657Z Sig Brng: 201
					SNR	Prob
Quarry Heights	2341	1104	079	259	-39	1.1E-20
Melbourne	0712	2136	210	030	---	0
Navarin Mys	0945	1430	357	177	-10	4.8E-09
Stalino	2136	0134	337	157	-328	0
Gardner Island	0538	1747	175	355	28	3.7E-03
Oust-Kiakhta	1307	2051	315	135	-102	1.3E-65
San Diego acft	0300	1245	069	249	9	1.8E-03
Los Angeles acft	0309	1245	067	247	9	1.6E-03
San Francisco acft	0335	1253	063	243	1	1.5E-02
Seattle acft	0408	1219	049	229	14	1.2E-02
Billings	SSS	SSS	357	177	-11	6.9E-10

Analysis

Click on the map to open a much larger version in a new window.

1. The fact that this bearing is well off the bearing of Gardner Island does not necessarily rule out Gardner as the signal source. The short duration and weakness of the signal could account for the difference. The PAA DF system was designed for strong signals of several minutes duration.
2. The bearing of station 3AR in Melbourne was near the reported bearing, but the 3100 kHz harmonic of the 3AR signal could not be heard at Midway.
3. There were no known voice-capable signal sources on 3105 kHz near the reported bearing or its reciprocal.
4. The bearings of Navarin Mys and Billings were near the reciprocal of the bearing of Gardner, and one of those stations could have been the source, although the reception probabilities suggest otherwise. A signal from Navarin Mys was 770,000 times less likely to be heard than a signal from Gardner, and a signal from Billings was 5.3 million times less likely to be heard.
5. The Itasca did not transmit⁶⁵ between 0330Z and 0730Z, and thus could not have been the source.
6. This signal was heard 8 minutes after the 0630Z KGMB broadcast to Earhart. The time proximity of this signal to the KGMB broadcast suggests it was a response to the broadcast, and west coast sources can be ruled out under the rationale used in the analysis of Bearing 3.
7. The Midway report indicates that they did not listen to KGMB prior to 0830Z, suggesting that they did not have first-hand knowledge of the 0630Z broadcast.
8. As in the case of Bearing 3, the only known voice-capable transmitter on 3105 kHz fitting the parameters of this case is KHAQQ on the Earhart aircraft.

A bearing of 175 degrees on a signal on 3105 kHz described as “a strong carrier” and a “steady unmodulated carrier” that “continued for over two hours.” The Midway report further stated that this signal “proved to be some unidentified station probably in South America or Russia and was later definitely disregarded as a possibility.”

Table 9: Signal Data for Bearing 5

Source	Sun Set Z	Sun Rise Z	Source Brg	Recip Brg	Midway Sunset: 0650Z Sunrise: 1657Z Sig Brng: 175
					SNR	Prob
Quarry Heights	2341	1104	079	259	-60	4.6E-78
Melbourne	0712	2136	210	030	---	0
Navarin Mys	0945	1430	357	177	-18	4.7E-06
Stalino	2136	0134	337	157	-217	0
Gardner Island	0538	1747	175	355	10	8.2E-06
Oust-Kiakhta	1307	2051	315	135	-9	4.8E-21
San Diego acft	0300	1245	069	249	-4	2.6E-05
Los Angeles acft	0309	1245	067	247	-4	7.9E-05
San Francisco acft	0335	1253	063	243	-1	9.7E-05
Seattle acft	0408	1219	049	229	0	1.2E-04
Billings	SSS	SSS	357	177	-16	3.9E-13

Analysis

Click on the map to open a much larger version in a new window.

1. This bearing line passes through Gardner Island.
2. A 2-hour transmission by KHAQQ would be plausible although it would have been necessary to keep the starboard engine running to avoid depleting the battery charge.
3. The Itasca did not transmit⁶⁶ on 3105 kHz between 0815Z and 1213Z, and thus could not have been the source of this signal.
4. There appears to have been a discrepancy between the clocks at Midway and on the Itasca. The Midway report⁶⁷ states that at 1235Z the Itasca was heard “on 3105 KC calling the Earhart plane giving instructions.” The radio log of the Itasca shows a transmission to Earhart at 1223Z requesting dashes. The next transmission to Earhart was at 1240Z and was logged as “CALLED EARHART FONE//.” If this is a condensed reference to a transmission requesting some action by Earhart, and was the signal that Midway reported hearing at 1235Z, then there was a 5 minute discrepancy between Midway and the Itasca. Unfortunately, there is not enough information available to resolve whether the error was in one or both of the clocks.
5. The PAA basis for deciding the source was an unidentified station in South America or Russia was not stated. But there were no known stations on 3105 in South America. There were 3 Russian stations near the reciprocal bearing: Stalino, Navarin-Mys, and Billings. The reception probabilities for signals from Stalino and Billings were too low for serious consideration. A signal from Navarin-Mys would be plausible, though very unlikely.
6. On balance, there is no clear weight of evidence for or against this signal originating at Gardner Island.

As noted earlier, the Itasca deployed an experimental direction finder on Howland Island, manned by Radioman Second Class Frank Cipriani. Leo Bellarts, the Itasca’s chief radioman in July 1937, described⁶⁸ this direction finder as a portable Navy unit using a loop antenna, that had neither slip rings connecting the loop to the output signal wires, nor limit stops to prevent over-twisting the signal wires. According to Bellarts, the wires were badly twisted and broken when the unit was returned to the Itasca.

The Howland station log entry⁶⁹ for 1105Z on July 5 stated:

Weak carrier on 3105. No call given. Unilateral bearing impossible due to night effect. Using small pocket compass to determine relative direction. Bearing only approximate SSE or NNW.

Howland sent a message⁷⁰ to the Itasca at 1455Z stating:

At 0035 HST, obtained bearing on a continuous wave of unknown origin indicating south southeast or north northwest on magnetic compass. Unable to obtain unilateral bearing due to night effect. No call given. Frequency is slightly above 3105 KCS.

Table 10: Potential Signal Sources for Bearing 6

Source	Sun Set Z	Sun Rise Z	Brg	Recip Brg	Howland Sunset: 0556Z Sunrise: 1746Z Sig Brng: SSE/NNW
					SNR	Prob
Quarry Heights	2341	1104	081	261	-71	9.1E-74
Melbourne	0712	2136	218	038	-15	5.8E-20
Navarin Mys	0945	1430	358	178	-38	1.1E-10
Stalino	2136	0134	333	153	-260	0
Gardner Island	0538	1747	157	337	42	5.8E-01
Oust-Kiakhta	1307	2051	321	141	-68	4.2E-70
Caribbean Petr acft	2312	1029	079	259	-111	1.3E-137
San Diego acft	0300	1245	054	234	-14	4.2E-06
Los Angeles acft	0309	1245	052	232	-14	1.1E-05
San Francisco acft	0335	1253	047	227	-14	1.2E-06
Seattle acft	0408	1219	037	217	-16	2.3E-05
Billings	SSS	SSS	357	177	-36	5.7E-25

Analysis

Click on the map to open a much larger version in a new window.

1. 0035 HST was 1105Z the same day.
2. The radio message to the Itasca paraphrased the Howland log entry, omitting the key facts that the signal was weak, and the compass used was a small pocket compass.
3. The computed reception probability for a signal from Gardner Island at the time was nearly 100 percent, contradicting Cipriani’s description of the signal strength, if the signal he heard was in fact from Gardner. However, a strong signal from Gardner could have been perceived as weak if the DF unit’s wires were already twisted and broken when the signal was heard.
4. It is not surprising that Cipriani was unable to get a unilateral bearing. The unsuitability of a loop antenna for direction finding of skywave signals was discussed earlier.
5. The Itasca was not the source of this signal. The ship did not transmit⁷¹ on 3105 kHz between 0815Z and 1213Z.
6. The lack of a stated signal duration precludes direct comparison with the signal reportedly heard at the same time by Midway in the Bearing 5 case.
7. Converting a magnetic compass bearing to a true bearing requires knowing the local magnetic variation and the magnetic deviation of the compass. The local variation might have been available from a navigational chart. But magnetic deviation depends on the mass and specific location of ferrous metal near the compass, and it is highly doubtful that Cipriani’s pocket compass had been calibrated for deviation. Therefore, the bearing axis reported by Cipriani must not be considered as only a very general approximation.
8. The potential signal sources generally northwest of Howland were Navarin Mys, Stalino, Oust-Kiakhta, and Billings. With the exception of Navarin-Mys, the reception probabilities for signals from those sites are too low for serious consideration. Navarin-Mys is plausible, but highly unlikely.
9. With the exception of Gardner Island, there were no known sources generally southeastward of Howland.
10. The reception probability for a signal from Gardner was 58 percent.
11. Given the available evidence, Gardner Island is the most plausible source of this signal.