Petition for review granted, order vacated, and case remanded for further consideration consistent with this published opinion. Judge GREGORY wrote the majority opinion, in which Judge KING joined. Judge AGEE wrote a dissenting opinion.
GREGORY, Circuit Judge:
Tumaini Temu is a Tanzanian national who suffers from severe bipolar disorder. In his home country, Mr. Temu was tortured by nurses and prison guards because of his illness. After entering the United States, he applied for asylum, arguing that he was persecuted because of his membership in a particular social group. The Board of Immigration Appeals (“BIA”) denied his application, finding that Mr. Temu was not a member of a social group under the Immigration and Nationality Act (“INA”), and even if he was, Mr. Temu did not show that he was persecuted because of membership in this group. Because we agree with Mr. Temu that the BIA’s opinion rests on factual and legal errors, we grant Mr. Temu’s petition for review, vacate the BIA’s order, and remand for further proceedings consistent with this opinion.
I.
The facts presented below are based on Mr. Temu’s testimony, as well as testimony from two expert witnesses who discussed Mr. Temu’s diagnosis and the conditions that individuals with mental illness face in Tanzania. The IJ credited the testimony of all three witnesses, and neither the BIA nor the government dispute any of the facts presented.
Mr. Temu’s troubles began during his final year at the University of Dares Salaam, when his mother died in a car accident. This spurred a mental breakdown that forced Mr. Temu to leave school, and he experienced a series of similar episodes that were later diagnosed as manifestations of bipolar disorder. During his manic episodes, Mr. Temu believes he has su[890]*890perhuman powers. He is visibly erratic and often walks into busy intersections to direct traffic because he thinks he has the ability to prevent car accidents. This behavior caught the attention of Tanzanian officials who took him to Muhimbili Hospital in Dares Salaam, Tanzania, in 2003.
Mr. Temu’s admission to Muhimbili Hospital kicked off years in asylums and prisons during which Mr. Temu suffered violent physical abuse. At his asylum hearing, an expert witness testified that Tanzanians consider mental illness to be shameful. In Tanzanian culture, severe mental illness with visibly erratic behavior is seen as a manifestation of demonic possession. Tanzanians even have a label for the group, referring to those with visibly severe mental illness as “mwenda wazi-mu,” which means demon-possessed. The expert witness testified that even medical professionals in Tanzania believe that severe mental illness accompanied by erratic behavior is caused by demonic possession. Laymen and doctors alike believe that demonic possession is contagious. For this reason, even though friends and family visited Mr. Temu during his first hospitalization, they deserted him within months.
The nurses at Muhimbili Hospital treated Mr. Temu with violence and abuse. Nurses tied Mr. Temu’s hands and feet for five to seven hours a day, four days per week. When Mr. Temu’s condition worsened, his “treatment” became more inhumane, as he was bound and beaten with leather straps for eight hours per day, five or six days per week. Hospital stints turned into prison stints, and the abuse continued. Prison guards beat Mr. Temu with a club about his elbows and feet four days per week. The beatings were so severe that he could not walk.
The record is unequivocal about what motivated the nurses’ and guards’ behavior. Throughout all his hospitalizations, the nurses referred to Mr. Temu as “mwenda wazimu.” The record also shows that while binding Mr. Temu and beating him with leather straps, the nurses said on multiple occasions, “this is how we treat people who are mentally ill like you.” J.A. 135. In prison, the guards also referred to Mr. Temu as “mwenda wazimu.” All prisoners were beaten, but Mr. Temu received worse beatings. However, other prisoners who also suffered from severe mental illness were beaten as much as Mr. Temu.
Upon coming to the United States, Mr. Temu applied for asylum, withholding of removal, and relief under the Convention Against Torture (“CAT”). 8 U.S.C. §§ 1158(b)(1)(A), 1231(b)(3)(A); 8 C.F.R. § 208.16. Mr. Temu argued that under 8 U.S.C. § 1101(a)(42), he faced severe persecution because of his membership in the social group of individuals with bipolar disorder who exhibit erratic behavior. The immigration judge (“IJ”) denied Mr. Temu’s asylum and withholding claims. In a finding adopted by the BIA, the IJ concluded that Mr. Temu’s proposed group lacks the elements of immutability, particularity and social visibility necessary to qualify as a particular social group under the INA. In addition, both the IJ and BIA concluded that even accepting Mr. Temu’s proposed group, he did not show that he was persecuted because of his membership in this group. However, the IJ granted Mr. Temu CAT relief. In doing so, the IJ and BIA found that Mr. Temu was tortured by nurses and prison guards because he was mentally ill.
Mr. Temu filed a timely appeal of the BIA’s decision, arguing that it committed error in denying him asylum and withholding of removal. We have jurisdiction to hear his case under 8 U.S.C. § 1252.
II.
Individuals qualify for asylum if they were persecuted “on account of ... [891]*891membership in a particular social group.” 8 U.S.C. § 1101(a)(42)(A).1 This appeal raises two questions. First, we must analyze whether Mr. Temu’s proposed group of “individuals with bipolar disorder who exhibit erratic behavior” qualifies as a “particular social group.” Second, we ask whether Mr. Temu was persecuted because of membership in his proposed group. Under Chevron, we give deference to the BIA’s interpretation of the phrase “particular social group.” See Cervantes v. Holder, 597 F.3d 229, 232 (4th Cir.2010) (citing Chevron U.S.A., Inc. v. NRDC, Inc., 467 U.S. 837, 844, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984)). However, in reviewing whether a group meets the BIA’s definition of “particular social group,” we overturn a denial of asylum if it is “manifestly contrary to the law and an abuse of discretion.” 8 U.S.C. § 1252(b)(4)(D); see Zelaya v. Holder, 668 F.3d 159, 165 (4th Cir.2012); Crespin-Valladares v. Holder, 632 F.3d 117, 124-126 (4th Cir.2011). We uphold factual findings unless no rational factfinder could agree with the BIA’s position. Crespin-Valladares, 632 F.3d at 124.
III.
We first consider the BIA’s conclusion that Mr. Temu was not persecuted because of membership in his proposed group. Because this is a factual finding, our task is not to decide how we would rule in the first instance. Rather, we must uphold the BIA’s finding unless no rational factfinder could reach the same conclusion. See Crespin-Valladares, 632 F.3d at 124. In spite of this stringent standard of review, we are compelled to vacate because the BIA’s finding on nexus contains two logical contradictions that no rational fact-finder could hold.
First, it is impossible to square the BIA’s conclusion with the undisputed facts of the case. The BIA credited Mr. Temu’s testimony in its entirety, J.A. 151, and he testified not only that nurses beat and bound him, but also that they explicitly told him that “[t]his is how we treat people who are mentally ill like you.” J.A. 135. Mr. Temu testified that in prison, the guards beat all prisoners, but Mr. Temu was singled out for worse beatings, and other prisoners with mental illness were beaten as much as Mr. Temu. J.A. 137. Throughout his time in prisons and hospitals, the nurses and guards referred to him as “mwenda wazimu.” J.A. 135-37. We fail to see how a rational factfinder could simultaneously credit these facts and also conclude that Mr. Temu was not persecuted because of his mental illness and its manifestations. It is difficult to imagine evidence that is more persuasive and unequivocal than a persecutor directly telling a victim, “[t]his is how we treat mentally ill people like you.” J.A. 135.
Second, the BIA’s nexus finding and CAT finding are at logical loggerheads. The BIA adopted the IJ’s finding that “there is no nexus between the respondent’s mistreatment and his defined particular social group, which is defined in part by bipolar disorder.” J.A. 74. That is, even accepting Mr. Temu’s proposed group, the BIA concluded that his beatings were due to his erratic behavior, not his bipolar disorder per se. Mere pages later, [892]*892however, the IJ granted CAT relief, finding that Mr. Temu “was singled out for more frequent beatings because he was mentally ill.” J.A. 156.2 We struggle to see how a rational factfinder could conclude both that Mr. Temu was not persecuted because of his membership in the group of individuals with bipolar disorder who exhibit erratic behavior, and also that he was singled out for beatings because of his mental illness. It might be possible to reconcile these conflicting findings, but it would demand logical acrobatics, and the BIA makes no attempt to explain how it can believe that Mr. Temu was not persecuted because of his bipolar disorder but was tortured because he was mentally ill.
We are mindful that reviewing courts should not substitute their own judgment for the BIA’s in areas where the BIA is entrusted with the power to adjudicate claims in the first instance. See Zuh, 547 F.3d at 504. This is not a case of a mere difference in judgment. When the very core of an opinion is internally contradictory and advances diametrically opposed conclusions within paragraphs, this is the very essence of irrationality. Because the BIA’s nexus finding collapses under the weight of its logical defects, we are compelled to vacate the BIA’s finding.
IV.
We next consider the BIA’s conclusion that Mr. Temu’s proposed group does not qualify as a “particular social group” under 8 U.S.C. § 1101(a)(42)(A). Through its case law, the BIA has formulated a three-part test for what constitutes a “particular social group.” See In re S-E-G-, 24 I. & N. Dec. 579 (BIA July 30, 2008). First, individuals in the group must “share a common, immutable characteristic ... that members of the group either cannot change, or should not be required to change.” Id. at 582-83 (internal citation and quotation omitted). In addition, the group must have social visibility, which means “the group should generally be recognizable by others in the community.” Id. at 586. Finally, the group must be defined with particularity, which means the group must have concrete, identifiable boundaries that allow an observer to distinguish members of a group from nonmembers. Id. at 584.
We must uphold the BIA’s conclusion that Mr. Temu’s group does not qualify as a particular social group unless it is “manifestly contrary to the law and an abuse of discretion.” Zelaya, 668 F.3d at 165. Because the BIA’s opinion rests on legal error, we must reverse. It is unclear from the BIA’s opinion whether it misapplied its own standard or applied a new standard without explanation, but in either ease, the BIA’s legal analysis is manifestly contrary to the law. Further, the BIA’s opinion rests on factual error.
A.
We first consider the BIA’s social visibility analysis. Social visibility does not mean ocular visibility: a group can qualify as a social group even if one cannot identify members of the group by sight. See, e.g., Henriquez-Rivas v. Holder, 707 F.3d 1081, 1087-88 (9th Cir.2013); Rivera-Barrientos v. Holder, 666 F.3d 641, 652 (10th Cir.2012). Rather, social [893]*893visibility speaks to whether a group is in fact recognized as a group. See In re CA-, 23 I. & N. Dec. 951, 959 (BIA 2006) (defining social visibility as whether a group is “understood by others to constitute” a social group). For example, in the United States, “Vietnam veterans, ... cancer survivors, blind people, Cajuns, practitioners of Falun Gong and hippies” would likely be identified as social groups, whereas “second-born children and haters of broccoli” would not. Henriquez-Rivas, 707 F.3d at 1096-97 (Kozinski, J., dissenting). Thus, many groups have qualified as socially visible under BIA case law, even though their members are not visibly identifiable. See In re C-A-, 23 I. & N. Dec. at 959 (citing groups defined by kinship ties, prior employment, and genital mutilation).
Mr. Temu’s group appears to have a strong case for social visibility, but the BIA never applied a permissible legal standard. The BIA found that while “Tanzanian society unquestionably targets individuals who exhibit erratic behavior for serious forms of mistreatment,” this mistreatment is not “limited to those who have a diagnosis of bipolar disorder.” J.A. 153. For example, the IJ noted that a visibly intoxicated person might exhibit erratic behavior and get targeted for mistreatment. Id. On its face, it might appear that the IJ’s opinion conflates the nexus requirement with social visibility, but in fact, the IJ’s argument is much more subtle. The IJ’s argument is that a lack of nexus is evidence of a lack of social visibility. Thus, if persecutors torture a wide swath of victims indiscriminately, this not only suggests a lack of nexus, but it also suggests that the persecutors did not even consider any one victim’s particular social group. This, in turn, suggests a lack of social visibility. In sum, because the persecutors used erratic behavior as an over-broad proxy for identifying victims, the persecutors did not view Mr. Temu’s proposed group as a group in the first place.
This conclusion does not show that Mr. Temu’s group lacks social visibility: it shows that Mr. Temu’s group lacks 20/20 visibility. The record is clear that Tanzanians view those with severe, chronic mental illness who exhibit erratic behavior as a group, since these individuals are singled out for abuse in hospitals and prisons and are specifically labeled “mwenda wazimu.” J.A. 137, 145. The nurses in this case explicitly said that “this is how we treat people who are mentally ill like you.” J.A. 135 (emphasis added). The fact that Tanzanians are overbroad in assigning this label to individuals does not show that social visibility is lacking. Though the persecution can be poorly aimed in theory, Tanzanians still appear to view the “mwen-da wazimu” as a group, and that is all that social visibility requires. See Henriquez-Rivas, 707 F.3d at 1089.
Another formulation of the social visibility test lends further support to this point. Circuit courts and the BIA have argued that a group is socially visible if it can show that it is singled out for worse treatment than other groups. For example, the BIA rejected a group of non-criminal informants as lacking social visibility because “informants are not in a substantially different situation from anyone who has crossed [the persecutors] or who is perceived to be a threat to the [persecutors’] interests.” In re C-A-, 23 I. & N. Dec. at 960-61; see also In re A-M-E & J-G-U-, 24 I. & N. Dec. 69, 75 (BIA 2007); Ramos-Lopez v. Holder, 563 F.3d 855, 861 (9th Cir.2009). In this case, Mr. Temu meets that test easily. The undisputed facts show that even though all prisoners were abused, Mr. Temu was singled out for worse abuse, with the exception of other prisoners with mental illness, who [894]*894received the same increased abuse as Mr. Temu.
This formulation of social visibility also illustrates the BIA’s legal misstep. Evidence that persecutors target an entire population indiscriminately can be evidence of no social visibility. Id. In that situation, the fact that members of a particular social group get caught in the same net is irrelevant. The BIA extended this reasoning to conclude that any time a persecutor’s net is too large, social visibility must be lacking. The folly of this legal conclusion can be demonstrated with a hypothetical. Imagine that an anti-Semitic government decides to massacre any Jewish citizens. Now, imagine that in putting its policy into practice, the government collects a list of surnames of individuals who are known to be Jewish and then kills anyone with the same surname. Jews and Gentiles alike might be murdered, but this does not change the fact that Jews have social visibility as a group. Meanwhile, under the BIA’s reasoning, the fact that persecutors might lump non-group members with group members is, by itself, enough evidence to find a lack of social visibility.
Similarly, an analogy to a group that qualifies as a particular social group is helpful in illustrating why the BIA’s analysis in this case is impermissible. There is no doubt under BIA or federal case law that kinship ties can serve as the basis for a particular social group. The BIA has identified “kinship ties” as a paradigmatic example of a particular social group. See Crespin-Valladares, 632 F.3d at 124-25 (citing In re C-A- 23 I. & N. Dec. at 959; In re H- 21 I. & N. Dec. 337, 342 (BIA 1996)). This Court and “every circuit court to have considered the question” have reached the same conclusion. Cres-pin-Valladares, 632 F.3d at 125 (collecting case law). Yet, under the BIA’s reasoning in this case, if persecutors were using a distinctive family trait like curly red hair to identify and persecute individuals, then family ties would not qualify as a particular social group, since persecution would not be “limited to those” who are in the group. J.A.' 153. These examples illustrate why the BIA’s application of its social visibility test is legally erroneous in this case. Requiring what amounts to 20/20 visibility, rather than social visibility, would lead to absurd conclusions that flout the case law of this Court, other circuit courts, and the BIA itself.
There is no mechanical way to separate “haters of broccoli” from “Vietnam veterans,” Henriquez-Rivas, 707 F.3d at 1096-97, but one highly relevant factor is if the applicant’s group is singled out for greater persecution than the population as a whole. In re C-A-, 23 I. & N. Dec. at 960-61. Similarly, evidence that a proposed group has a specific label in a society is highly relevant. A group cannot be defined solely by the fact of its persecution, Gatimi v. Holder, 578 F.3d 611, 616 (7th Cir.2009), so evidence that members of a society have a label for a proposed group helps suggest that the group has a common thread outside of its victimhood, assuming of course that the label is not something like “persecution victims.” In sum, we vacate the BIA’s social visibility finding because it rests on legal error.
B.
The BIA also commits legal error in concluding that Mr. Temu’s group lacks particularity. Specifically, the BIA erred because it broke down Mr. Temu’s proposed group into pieces and rejected each piece, rather than analyzing his group as a whole. Once again, the BIA applied an impermissible legal standard because it rejected groups that Mr. Temu never proposed.
[895]*895A social group must have identifiable boundaries to meet the BIA’s particularity element. For example, the group “affluent Guatemalans” fails because the group changes dramatically based on who defines it. See In re A-M-E-, 24 I. & N. Dec. at 76. Affluent might include the wealthiest 1% of Guatemalans, or it might include the wealthiest 20%. Therefore, this group lacks boundaries that are fixed enough to qualify as a particular social group.
In this case, the BIA found no particularity because bipolar disorder is too broad and erratic behavior is too fuzzy. First, bipolar disorder covers a wide range of severity. At its least severe, the disorder can be so mild as to be outwardly undetectable. Therefore, the disorder covers too broad a spectrum of behavior to have identifiable boundaries. The other component of Mr. Temu’s proposed group is erratic behavior, but this, too, lacks particularity. The definition of erratic behavior changes based on who defines it, and it is difficult to put precise, identifiable boundaries on what constitutes erratic behavior. Because each part of Mr. Temu’s proposed group lacks particularity, the BIA concluded that the group as a whole fails.
The BIA’s opinion commits legal error by splitting Mr. Temu’s group in two and rejecting each part, rather than considering it as a whole. See Crespin-Valladares, 632 F.3d at 125. The BIA is correct that the label of mental illness can cover a broad range of severity. On its own, it is possible — though we do not decide — that the group of individuals with bipolar disorder lacks particularity because of its breadth, but that is not Mr. Temu’s proposed group. Rather, Mr. Temu limits his group to those individuals with bipolar disorder who exhibit outwardly erratic behavior. It may well be that mental illness lacks particular boundaries, since the label covers a huge swath of illness that ranges from life-ending to innocuous. Mr. Temu’s group does not suffer from the same shortcoming, because it is limited to a specific mental illness so severe that individuals are visibly, identifiably disturbed.
Similarly, the BIA rejects erratic behavior as lacking particularity. Erratic behavior is difficult to define and subjective. We doubt that “individuals who exhibit erratic behavior” would qualify as a particular social group, but again, Mr. Temu proposed no such group. Rather, Mr. Temu’s group is limited to individuals who exhibit erratic behavior and suffer from bipolar disorder. Unlike “erratic behavior,” the term bipolar disorder has well-defined, identifiable characteristics. See generally American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders (5th Ed., 2013); World Health Organization, International Statistical Classification of Diseases and Related Health Problems V(F30)-(F39) (10th ed., 2010). The BIA faulted Mr. Temu’s group because it lacks an “adequate benchmark,” J.A. 4, but that is precisely what the DSM-V supplies with regard to the other component of Mr. Temu’s group. J.A. 4. Thus, erratic behavior has unclear boundaries that the other component of Mr. Temu’s group supplies. In turn, bipolar disorder covers a broad spectrum of behavior that is sharply limited by the requirement of erratic behavior.
In essence, the BIA committed legal error because it missed the forest for the trees. While each component of Mr. Temu’s group might not satisfy the particularity requirement individually, the BIA must consider Mr. Temu’s definition as a whole. See Crespin-Valladares, 632 F.3d at 125 (“[The BIA’s legal error] flowed from the fact that, as the Government concedes, the BIA’s removal order rejected a group different from that which the [applicants] proposed”). For example, we [896]*896have recently found that the “group consisting of family members of those who actively oppose gangs in El Salvador by agreeing to be prosecutorial witnesses” qualifies as a particular social group. Id. at 120-121, 125-26. Each component of the group in Crespin-Valladares might not have particular boundaries. “Prosecutorial witnesses” might reach too broad a swath of individuals; “those who actively oppose gangs” might be too fuzzy a label for a group. Our case law is clear, however, that the group as a whole qualifies. In this case, the BIA took issue with the component parts of Mr. Temu’s group, but it never reached the stage of assessing the particularity of Mr. Temu’s group as a whole. Instead, it considered and rejected two different groups that were based on pieces of Mr. Temu’s group.
Thus, the BIA’s particularity analysis was based on legal error. The INA requires that an individual be persecuted because of membership in a “particular social group.” 8 U.S.C. § 1101(a)(42). Nothing in the statute requires that if a group is defined by a collection of traits, that each individual trait must meet all the criteria for a “particular social group.” Time and again, case law from this Court, other circuits, and the BIA has accepted social groups that, as part of their definitions, contain components that might not meet the BIA’s legal standards. See Crespin-Valladares, 632 F.3d at 120-21 (accepting the group of “family members of those who actively oppose gangs in El Salvador by agreeing to be prosecutorial witnesses”); Tapiero die Orejuela v. Gonzales, 423 F.3d 666, 672 (7th Cir.2005) (accepting the group of “educated, landowning class of cattle farmers”); In re CA- 23 I. & N. Dec. at 960 (citing favorably the group of “young women of a particular tribe who were opposed to female genital mutilation” as a particular social group) (citing In re Kasinga, 21 I. & N. Dec. 357, 365-66 (BIA 1996)). Notably, the BIA itself has accepted individuals with bipolar disorder as a particular social group in the past, albeit in cases that, like this one, were unpublished. In re Daniel Francisco Lopez-Sanchez, 2010 Immig. Rptr. LEXIS 7882 (BIA 2010); In re A-, (BIA May 31, 2007) (slip op.). In fact, after oral arguments in this case, the BIA issued a decision accepting the particular social group of individuals in Ghana with severe mental illness, specifically bipolar disorder, who are indigent and lack family support. In re — , (BIA Nov. 15, 2013) (slip op.). These cases illustrate that in making asylum determinations, the BIA must consider an individual’s proposed group as a whole. Once again, it is unclear whether the BIA misapplied its own legal standard or advanced a new legal standard that involves piecemeal analysis. Either way, the BIA’s opinion is “manifestly contrary to the law.” Crespina-Valladares, 632 F.3d at 126.
C.
Particular social groups must also be characterized by immutability, and Mr. Temu’s proposed group easily satisfies this final element. The BIA’s conclusion to the contrary rests on factual error. The BIA opinion finds that “there is no cure for bipolar disorder,” J.A. 154, so there is no doubt that bipolar disorder is immutable. However, the BIA found no immutability because Mr. Temu’s erratic behavior can be controlled with medication. Id. However, in the same opinion, the BIA also adopts the finding that “there is no consistent access to the medications the respondent needs in Tanzania,” and that because his family has abandoned Mr. Temu, he will not be able to obtain what medications are available. J.A. 146. Once again, the BIA’s opinion advances two factual findings that are impossible to reconcile with[897]*897out violating fundamental rules of logic. According to the BIA, Mr. Temu’s disorder will never be cured and will only worsen. J.A. 146. He can only control his behavior with medication, but he will not have access to this medication in Tanzania. The inescapable conclusion from this finding is that if he is returned to Tanzania, Mr. Temu will not be able to control his behavior. In sum, Mr. Temu’s membership in his proposed group is not something he has the power to change.
The BIA’s position has been explicitly rejected by the Seventh Circuit and by the BIA itself. In Kholyavskiy v. Mukasey, the Seventh Circuit considered an asylum claim based in part on the applicant’s mental illness. 540 F.3d 555, 572-74 (7th Cir.2008). The BIA found no immutability because even if the disease is incurable, the individual’s behavior could be controlled through medication — medication that the applicant would not have access to if returned to his home country. The Seventh Circuit concluded that the BIA’s immutability argument, which is identical to the one presented here, had no factual basis. Id. at 573. Further, the BIA itself has found that severe mental illness is immutable in two unpublished opinions, explicitly ruling that “bipolar disorder [is] a chronic psychiatric condition subject to treatment but not cure, and thus it [is] an immutable characteristic.” In re A-, (BIA May 31, 2007) (slip op.); In re — , (BIA Nov. 15, 2013) (slip op.) (finding immutability because bipolar illness is permanent regardless of medication). These cases reach the same conclusion that is compelled by the facts of this case: when an individual suffers from an incurable mental illness, it is of no relevance that somewhere in the world, there exists medication that can help him control the illness. If he cannot access the medication, his behavior is as effectively immutable as if the medication did not exist. Further, the underlying bipolar disorder will never change. While it can be managed, this does not mean that it can be cured. These facts compel the finding that Mr. Temu’s group membership is immutable.
V.
For the foregoing reasons, we grant the petition for review, vacate the BIA’s order affirming the IJ’s decision, and remand for further consideration consistent with this opinion.
PETITION FOR REVIEW GRANTED, ORDER VACATED, AND CASE REMANDED.