The details of the attack show that the FNL was the chief force in the slaughter at Gatumba. Witnesses both in and near the camp agree that the attackers arrived making music and singing religious songs in Kirundi. This has been standard practice for FNL attacks for several years, a practice not found among other Burundian armed groups nor ordinarily among groups in the Congo. Many witnesses said that women accompanied the combatants and carried off looted goods. Several witnesses commented also on the young age of some of the attackers. In the last two years, FNL forces included women and children in many attacks. In addition, the site of the refugee camp at Gatumba lies near the Rukoko forest where the FNL are known to have an important base. 71
A twenty-five-year-old FNL combatant arrested by Burundian authorities on August 19 confessed to participating in the Gatumba attack, saying that he himself killed some of the refugees. The circumstances of his arrest were peculiar: he had supposedly been stopped by a group of young men in charge of security in the neighborhood of Ngagara, a very heavily Tutsi area and presumably not a place where a FNL combatant would feel at ease. The security detail invited him to share a few beers, after which he confessed his crime to them. Although the circumstances of his arrest raise considerable doubt, some UN investigators found some of his information credible. He was said to have been able to explain, for example, where and how some of the group crossed the border. Authorities have not made public the full extent of his information nor have they revealed where he is being detained.72
The FNL has accepted responsibility for the attack. Pasteur Habimana, spokesman for the movement, was the first to make such a statement. Early on the morning of the attack he called several Burundian journalists to castigate them for having broadcast reports that the perpetrators of the massacre had come from the Congo and were mostly Rwandan rebels and Mai Mai.73
Even after it became increasingly clear that accepting responsibility for the attack might be seriously damaging to his group, Habimana as well as the national secretariat made no retraction. They did elaborate on several reasons why the FNL had made the attack. They referred to the many killings of civilians that had gone unpunished in the years of conflict in Burundi, seeming to suggest that attacking the Banyamulenge was a justifiable response to these previous killings. Habimana also claimed that the FNL force had attacked the military and police camps and had pursued soldiers and police who fled from their camps to the refugee camp. There, said Habimana, Banyamulenge had brought out arms that had been hidden and fired on the FNL. As the days passed, he elaborated this explanation to the point of calling the refugee camp the Banyamulenge general staff headquarters. No evidence supports these claims.74
Habimanas declaration that the FNL was responsible for the massacre, improbable though his justifications seem, appears to be correct.
Information about languages spoken by the attackers cannot by itself identify their ethnicity or nationality, far less their affiliation with any particular group or army. Many persons in the regionparticularly those affiliated with armed bandshave lived outside their home regions and at least some of them have mastered the languages of the areas in which they lived. Thousands of Burundians, for example, have lived in parts of Fizi where Kibembe is spoken and some of them speak Kibembe. In addition, many people from Burundi, Congo, and Rwanda speak Swahili, whether they have ever traveled abroad or not.
When testimony about language is put together with other information, however, it can assist in identifying the attackers. All survivors interviewed by Human Rights Watch researchers report having heard attackers speak Kirundi. Burundian witnesses near but not in the camp say the same thing. The songs and bits of phrases reported by many witnesses are all Kirundi. 75 No other language was cited by witnesses as often or as consistently. This information strengthens the conclusion that FNL forces played the most important role in the attack.
Some survivors say they also heard Kinyarwanda, Swahili, Kifulero, Kibembe, and Lingala but when asked for the exact words they heard, few could provide convincing details. In one exceptional case, the young Mubembe mother mentioned above said that her conversation with an attacker was conducted in Kibembe, but according to a second witness, another attacker at the same tent appeared not to understand Kibembe. The attacker asked the occupants of the tent in Kirundi, Who are you? When people replied in Kibembe, he seemed not to understand their answers and shot into the group.76
One survivor who had been a RCD-Goma intelligence agent in Uvira spoke often to press and investigators; he presented several variants of how his life had been spared, but all involved his speaking Kifulero and being understood by attackers who also knew that language. His account became so well known that other survivors who were asked for evidence of languages spoken sometimes cited his experience when they had no specifics of their own to offer.77
Considering witness testimony, the extent, intensity and duration of the attack, and the well-armed nature of the attackers, it seems likely that fewer than one hundred attackers carried out the massacre of the refugees. As mentioned above, witnesses say that only one or two, sometimes three attackers came to the entry of each tent. The tents were not attacked simultaneously, meaning that some attackers struck in one place before moving on to another during the course of the hour they were there. Nor was the attack of uniform intensity: of the fifteen tents, those at either end of the rows were hardest hit while some tents escaped relatively unscathed. Attackers approached from one direction and did not surround the tents, although the structures occupied a relatively small terrain. Because of this, some people were able to escape from the back of the tents and then to flee to safety in the surrounding bush. The figure of approximately one hundred attackers also accords generally with estimates from Burundians resident in the immediate area.78
The combatants who attacked the military campand perhaps the police campalso appear to have been few in number, certainly not one or two companies suggested by the commander of the military camp. It is frequent FNL practice to attack military posts with small groups of five to ten men, who set up a barrage using a great deal of ammunition.
If the attackers numbered only a hundred or so, FNL forces could presumably have made up all or almost all of them. Evidence about languages spoken, although not conclusive, suggests there may have been some others present. In addition, witnesses recall attackers having said such phrases as Our comrades told us not to kill on the side of the camp with the Burundians, words which suggest that the speaker was not himself a Burundian.79 Current evidence does not permit further conclusions about the number or ethnic or political affiliation of such persons. Given the presence of Rwandan rebels in Burundi, some of them might have joined with the FNL, but their participation would not necessarily implicate the FDLR. Some Bembe are members of Mai Mai groups, so the presence of at least one speaker of Kibembe might suggest a link to the Mai Mai, but Human Rights Watch researchers found no other indication of Mai Mai involvement.
Both the FDLR and Congolese authorities, including Mai Mai, have denied any participation by troops under their command.80
 Human Rights Watch interviews, Gatumba and Bujumbura, August 17, 18, 19, 20, 21 and 22, 2004.
 Human Rights Watch interviews, Bujumbura, August 20, 21, and 26, 2004.
 Human Rights Watch interviews, Bujumbura, August 20 and 21, 2004.
 Human Rights Watch interviews, Bujumbura, August 20 and 21, 2004; FNL Secretariat National aux relations exterieurs, COMMUNIQUE PH-FNL 010/08/2004, dated August 30, 2004.
 Human Rights Watch interview, Gatumba, August 22, 2004.
 Human Rights Watch interview, Bujumbura, August 19, 2004.
The witness has two wives, only one of whom was of the Banyamulenge. One account is that he was spending the night with the non-Banyamulenge wife who was in a tent on the Burundian side of the camp and persuaded attackers that he was not Banyamulenge by speaking Kifulero to them. Human Rights Watch interviews, Gatumba, August 18 and Bujumbura, August 19, 2004.
 Human Rights Watch interview, Bujumbura, August 25, 2004.
 Human Rights Watch interviews, Bujumbura, August 21 and 24, 2004.
 Agence France Presse, Massacre de Gatumba : les rebelles rwandais démentent être impliqués, August 16, 2004.