dissenting:

Petitioner testified that he was arrested for hosting a party with his gypsy friends and that the police made disparaging remarks about gypsies when they arrested him. These facts support the majority’s conclusion that the arrest was on account of his gypsy ethnicity. But the beatings the majority believes amount to persecution happened after the arrest, and petitioner offered no evidence to connect that abuse to his ethnicity.

We have sometimes found derogatory comments sufficient to establish the motive for persecution, but only where they were made “[i]n the course of persecuting” the alien. Baballah v. Ashcroft, 367 F.3d 1067, 1077 (9th Cir.2004); see also, e.g., Maini v. INS, 212 F.3d 1167, 1176 (9th Cir.2000) (“[T]he ‘repeated beatings coupled with explicit expressions of [religious] hatred’ ... *732 compel a finding of persecution on account of religion.” (quoting Duarte de Guinac v. INS, 179 F.3d 1156, 1162 (9th Cir.1999)) (omission and first alteration added)). When someone is called names while being beaten, it’s reasonable to associate the two. Here, petitioner was abused in custody— after the arrest and, for all we know, by different officers. Petitioner has presented nothing to compel a finding that the earlier derogatory remarks were connected to the subsequent abuse.

The majority attempts to tie the improper arrest to the beating by suggesting that the abusers knew petitioner’s ethnicity, having been informed of that fact by the officers who arrested him. Maj. op. at 728. But this is hardly an inference that’s compelled by the record. Here’s the relevant excerpt from petitioner’s hearing before the IJ:

A.R. 84. And, in his credible fear interview, petitioner explained:

A.R. 203. There is no other evidence on this point.

Petitioner never testified that the officers who beat him while he was in prison were the same ones who arrested him. Nor did he say that the statement about initializing gypsy gatherings was made at the police station or otherwise in the presence of the officers who later beat him. And he admitted that he doesn’t look like a gypsy. See A.R. 102 (noting that people couldn’t tell petitioner was a gypsy just by looking at him).

The most we can say about the record is that it would support an inference that the officers who beat petitioner knew of his gypsy ethnicity — had the IJ made such a finding. But how can we possibly say that’s the only finding the record will support? Our job isn’t to nitpick the IJ’s findings, looking for ways we can construe the record to undermine them. Rather, we must uphold the IJ’s findings unless the record compels a contrary result. See Ochave v. INS, 254 F.3d 859, 862 (9th Cir.2001) (“When reviewing for substantial evidence, we must uphold the IJ’s findings unless the evidence not only supports, but compels, contrary findings.” (citing INS v. Elias-Zacarias, 502 U.S. 478 , 481 n. 1, 112 S.Ct. 812 , 117 L.Ed.2d 38 (1992))). Given the ambiguities in petitioner’s testimony, a reasonable fact-finder could have found that the arresting officers never informed their colleagues at the police station that petitioner is a gypsy, and that the officers who did the beating had no other basis for knowing petitioner’s ethnicity.

Yet, even if the majority’s interpretation of the record were the only reasonable one, showing that the officers at the station were aware of petitioner’s gypsy ethnicity hardly proves they abused him because of it. If beatings in Bulgarian prisons were rare, we might infer such a connection. But we can’t take this leap here because the IJ specifically found that any suspect held in custody by the Bulgarian police has a significant risk of being abused. See E.R. 59 (“Reports *733 continue that criminal suspects in police custody run a significant risk of being mistreated.”) (quoting Bureau of Democracy, Human Rights, & Labor, U.S. Dep’t of State, Bulgaria Country Report on Human Rights Practices for 1998 (1999) [hereinafter 1998 Human Rights Report ], reprinted in E.R. 252, 254). The State Department mentions, for instance, that “[sjecurity forces beat suspects and inmates and at times arbitrarily arrested and detained persons.” 1998 Human Rights Report at E.R. 252. In addition, “[ejredible sources reported cases of brutality committed by prison guards against inmates.” Id. at E.R. 255. Other sources report more of the same. See UNHCR Ctr. for Documentation & Research, Background Paper on Bulgarian Refugees and Asylum Seekers (1994), reprinted in E.R. 190, 198 [hereinafter UNHCR Background Paper ] (“While on a national level all citizens are granted constitutional protection against such arbitrary practices as illegal detention ... and cruel or inhuman treatment, incidents on the local level are known to occur.”); Human Rights Watch, World Report 1999: Bulgaria: The Role of the International Community, reprinted in E.R. 134, 134 (noting “ongoing abuses by the police and secret services”).

The IJ also found that abuse of prisoners “is not keyed to any ethnic or racial group,” E.R. 59, and this finding, too, is supported by substantial evidence. Both the State Department’s 1998 report on human rights in Bulgaria and a 1997 State Department report describe prisoner abuse, but neither suggests it is limited to minorities. See 1998 Human Rights Report at E.R. 254; Bureau of Democracy, Human Rights & Labor, U.S. Dep’t of State, Bulgaria: Profile of Asylum Claims and Country Conditions 2 (1997), reprinted in E.R. 112, 113 (“[TJhere remain problems ... in the treatment of detainees and prisoners.”). Although there are suggestions that gypsies are frequent targets of police beatings, see, e.g., Richard Groves, Changing the Face of Policing in Bulgaria, CPRSI Newsletter, Apr. 1996, reprinted in E.R. 180, 181 (describing “allegations of police brutality” toward gypsies); UNHCR Background Paper at E.R. 198 (noting reports that “police frequently are either the perpetrators of violence against Roma or they fail to intervene when attacks are instigated”), petitioner has offered nothing to suggest— let alone compel — a finding that gypsies in Bulgarian prisons are treated worse than anyone else, nor that he himself was treated worse than other inmates. Without such a finding, there’s no basis for jumping from the mere fact of abuse to the inference that it was on account of petitioner’s ethnicity.

My colleagues misunderstand my argument when they respond that “[ajsylum seekers who have fled from generally repressive regimes have no higher a burden than those who have fled from generally benign countries.” Maj. op. at 730. I’m not saying petitioner’s treatment wasn’t severe enough to be persecution; I assume it was. My point is that, since petitioner offered no evidence that he was singled out for mistreatment as a gypsy, he hasn’t shown that the abuse was on account of his ethnicity. Thus, no matter how bad or how frequent, his beatings don’t amount to persecution on one of the grounds specified in the asylum statute, if everyone else in his situation suffered exactly as he did.

The record does not compel the conclusion that the officers who abused petitioner at the police station knew he’s a gypsy. But even if it did, petitioner offered no evidence to show he was abused because of that ethnicity, and the IJ’s well-supported *734 finding that all Bulgarian prisoners are at significant risk of being abused precludes any such inference. Because an arrest alone does not compel a finding of persecution, see Prasad v. INS, 47 F.3d 336, 339-40 (9th Cir.1995), I would deny the petition for review and express no view as to any of the other points it raises.