David Eduardo Castañeda-Castillo ("Castañeda") entered the United States together with his family in August 1991 on a tourist visa, overstayed and in January 1993 applied for asylum. In July 1999, the responsible agency—then the Immigration and Naturalization Service ("INS")—began a removal proceeding, 8 U.S.C. § 1227(a)(1)(B) (2000). Thereafter, Castañeda testified at successive hearings seeking to establish his need for asylum and to refute a claim that he was barred from asylum by reason of having assisted or otherwise participated in persecution. Id. § 1158.

Castañeda, the only witness at the hearings, testified in substance that he had been commissioned as a lieutenant in the Peruvian military in 1983. In January 1985 he was transferred to an antiterrorist unit in an area designated an emergency zone in which the Shining Path was active. The Shining Path is a revolutionary Marxist organization, well known for its energy and brutality, that has been at war with the Peruvian government for many years.

In August 1985, the military conducted an operation to search for Shining Path members in the village of Llocllapampa in the Accomarca region. Two patrols, one of them headed by Sublieutenant Telmo Hurtado and the other by Lieutenant Riveri Rondon, were to enter the village and conduct the search. Two other patrols, one of which was led by Castañeda, were assigned to block escape routes from the village. Castañeda's patrol located itself about three to five miles from the village on either side of a path.

The first two patrols entered the village and there followed a brutal massacre of dozens of innocent villagers, including many women and children. Castañeda said that he was in radio contact with his base commander but not with the patrols entering the town, and that he did not know when the attack had occurred or that it had turned into a massacre of civilians. According to Castañeda, he did not learn of the massacre until about three weeks after the operation when he heard that Hurtado had confessed to executing civilians.

The Peruvian Senate Human Rights Commission conducted an investigation and concluded that Hurtado's patrol had been primarily responsible for the massacre. However, the report stated that Hurtado was "only a piece of a large picture and it is necessary to study whether he acted on virtue of expressed verbal orders or if he acted as such because that is how he has formed." As to Castañeda, the report concluded that his squad "was not involved in any confrontations with fugitive civilians."

Castañeda was charged before a military tribunal with homicide and abuse of authority, but the charges were dismissed. Hurtado, Riveri Rondon, and other officers were also charged, and Hurtado was convicted of abuse of authority, given a sentence which was later commuted in a general amnesty, and eventually restored to duty and promoted to captain.

After the court martial, Castañeda returned to duty and was promoted. However, he had been publicly associated with the massacre and (he testified) began to receive death threats from the Shining Path. According to Castañeda, the Shining Path had attempted to murder him on two occasions and to kidnap one of his daughters, and a neighbor and military colleague who had received similar threats was murdered by the Shining Path. These events, Castañeda said, prompted him to flee to the United States with his family.

Under the Immigration and Nationality Act, an alien is not eligible for asylum or withholding of removal if the Attorney General determines that "the alien ordered, incited, assisted, or otherwise participated in the persecution of any person on account of race, religion, nationality, membership in a particular social group, or political opinion." 8 U.S.C. § 1158(b)(2)(A)(i); see also id. § 1231(b)(3)(B)(i). In October 2004, following the series of hearings, the immigration judge ("IJ") found that this bar applied to Castañeda.

After a discussion of Castañeda's testimony, the IJ concluded that Castañeda was not credible; the IJ also said that even if Castañeda's testimony were believed, Castañeda had nevertheless assisted in persecution because the "objective effect" of his participation in the operation as a blocking force was to confine the villagers and thus aid in their massacre. In September 2005, the Board of Immigration Appeals ("BIA" or "Board") affirmed the IJ's ruling in a brief decision, agreeing that Castañeda had not been a credible witness and seemingly endorsing the "objective effect" ground as well.

Castañeda petitioned to this court for judicial review. A divided panel reversed the BIA and determined that the BIA's adverse credibility determination was not supported by substantial evidence, Castañeda-Castillo v. Gonzales, 464 F.3d 112 (1st Cir.2006), and that the persecutor bar did not apply to Castañeda. It remanded solely for a determination as to the merits of his asylum and withholding of removal application—a separate and sequential issue that is not before us. On petition by the government, this court agreed to rehear the case en banc, vacating the panel decision.

The Legal Issue. The statute that bars persecutors has a smooth surface beneath which lie a series of rocks. Among the problems are the nature of the acts and motivations that comprise persecution, the role of scienter, whether and when inaction may suffice, and the kind of connection with persecution by others that constitutes "assistance." The more one ponders the variety of possible situations, the less confident one becomes of a useful, all-embracing answer.

Yet the inquiry in this case can be narrowed. The government asserts and Castañeda does not deny that the deliberate massacre of civilians because of their perceived connection with or support for the Shining Path amounted to persecution. Nor does Castañeda dispute that, if he had been aware in advance of a plan to murder civilians, his role in leading a patrol to block escape routes could be treated as culpable assistance. We think that the same could be true if he learned of a massacre in progress and continued to assist by blocking escape routes.

What remains in dispute is the legal question whether the persecutor bar would apply to Castañeda if he had no prior or contemporaneous knowledge of the murder of civilians. Castañeda, saying that he lacked such knowledge, argues that the answer is no. By contrast, the government has taken the position, most fully elaborated in a letter from government counsel sent in response to a query from the panel, that such culpable knowledge is not required by the statute.

For this view the government offers several arguments. One—the lack of an explicit scienter requirement in the statute— is not persuasive: the term "persecution" strongly implies both scienter and illicit motivation. Dictionary definitions, as well as the Board's own precedent, bear this out.1 So does common sense: the bus driver who unwittingly ferries a killer to the site of a massacre can hardly be labeled a "persecutor," even if the objective effect of his actions was to aid the killer's secret plan.

The government cites Fedorenko v. United States, 449 U.S. 490, 101 S. Ct. 737, 66 L. Ed. 2d 686 (1981), Xie v. INS, 434 F.3d 136, 142-43 (2d Cir.2006), and In re Rodriguez-Majano, 19 I. & N. Dec. 811, 814-15 (BIA 1988), for the view that scienter is not required and that the "objective effect" of an alien's actions is all that matters. But these decisions merely say that one can "assist" in persecution even if his assistance is involuntary; the involuntariness claims seem to have fallen somewhere between a showing of true duress and an "obeying orders" defense. Cf. Hernandez v. Reno, 258 F.3d 806, 814-15 (8th Cir. 2001).

The government also cites various court and BIA precedents that suggest that the "totality of the relevant conduct" may be taken into account in determining whether one has assisted in the persecution of others. But these cases tend to reaffirm the need for some degree of moral culpability; and again, most focus on the question whether certain conduct constitutes "assistance," not on the issue of scienter.2

Two circuit cases might be read to depreciate the need for scienter, see United States v. Schmidt, 923 F.2d 1253, 1258 & n. 8 (7th Cir.), cert. denied, 502 U.S. 921, 112 S. Ct. 331, 116 L. Ed. 2d 272 (1991) (involving an SS officer who guarded a concentration camp and escorted inmates to and from forced labor sites); Kulle v. INS, 825 F.2d 1188, 1193 (7th Cir.1987), cert. denied, 484 U.S. 1042, 108 S. Ct. 773, 98 L. Ed. 2d 860 (1988) (same). But in both, knowledge of persecution was not in doubt and could be inferred from presence at the place of persecution. And most case law assumes or affirms the need for some degree of subjective fault. See note 2, above.

The government's better arguments turn on the need for some flexibility in applying the statute to gray areas and the latitude implicitly confided to the Attorney General in administering the scheme. Chevron U.S.A., Inc. v. Natural Res. Def. Council, 467 U.S. 837, 843-44, 104 S. Ct. 2778, 81 L. Ed. 2d 694 (1984). Consider, for example, cases of willful blindness or strong suspicions, or an abettor who knows generally of a pattern of persecution while being ignorant of specific incidents. There may well be gray-area cases where less than full and detailed knowledge may suffice.

But in the present case, it appears as if Castañeda either had guilty knowledge or (as he claims) knew nothing about the massacre until after it had occurred. If this is a case somewhere in between—say, of strong suspicions or willful blindness on Castañeda's part—the government has not spelled out such a possibility. No such scenario is described in the IJ or Board decisions, and the government does not point to any evidence for such an "in between" assessment.

Nor is it an excuse for dispensing with scienter that knowledge may be hard for the government to prove: in cases like ours, once the government introduced evidence of the applicant's association with persecution, it then became Castañeda's burden to disprove that he was engaged in persecution. 8 C.F.R. § 1208.13(c)(2)(ii); see Abdille v. Ashcroft, 242 F.3d 477, 491 (3d Cir.2001). Here, this would require Castañeda to disprove knowledge.3

We hold that presumptively the persecutor bar should be read not to apply to Castañeda if his version of his state of mind is accepted. On remand the agency can, if it wants, try to develop a construction more favorable to the government. But this would have to be done expressly and persuasively, and not by vague reference to the "totality of . . . conduct" that conflates the question whether one's conduct constitutes "assistance" with the question whether one possessed such scienter as may be required under the circumstances.

The IJ, Board, and government's filings can all be read to assume that scienter is not required and that any conduct with an "objective effect" of facilitating persecution by others would suffice to trigger the bar. The Board stated that "it may be true" that Castañeda "lacked knowledge regarding the events which took place following the August 8 massacre, including whether and how many people were killed." Although the IJ came closer to suggesting that Castañeda had guilty knowledge, the Board did not adopt this ground and the IJ himself assumed that such knowledge was unnecessary.

If the IJ and Board rested their decision upon a misunderstanding of the legal elements of persecution, the ordinary remedy is a remand to allow the matter to be considered anew under the proper legal standards. INS v. Orlando Ventura, 537 U.S. 12, 16-17, 123 S. Ct. 353, 154 L. Ed. 2d 272 (2002) (per curiam). The main question is how far this misunderstanding infected the fact-finder's conclusion, and the burden of showing lack of prejudice is normally on the party wishing to excuse an error.

This would be a different case if the evidence clearly established that Castañeda had guilty knowledge, or that his denial of knowledge was unworthy of belief, so that remanding "would be an idle and useless formality." NLRB v. Wyman-Gordon Co., 394 U.S. 759, 766 n. 6, 89 S. Ct. 1426, 22 L. Ed. 2d 709 (1969); see also Li v. INS, 453 F.3d 129, 137 (2d Cir.2006). Here, however, there is no direct evidence of Castañeda's knowledge; and the IJ's and the Board's adverse credibility findings are independently vulnerable.

Credibility. Because the burden rested on Castañeda to show that he was not a persecutor, the IJ and the Board needed to do no more than make a well-reasoned and specific determination based in the record that Castañeda was not credible in denying timely knowledge of the massacre. The rule in Moore v. Chesapeake & Ohio Ry. Co., 340 U.S. 573, 576, 71 S. Ct. 428, 95 L. Ed. 547 (1951)—that a fact cannot be established by disbelieving a witness's denial of that fact—does not assist one who (like Castañeda) bears the burden of proof as to the existence of the fact (here, lack of knowledge) and fails to carry it.

Although the fact-finder on the scene has the advantage as to demeanor evidence, Cordero-Trejo v. INS, 40 F.3d 482, 487 (1st Cir.1994), judges are expected— unlike juries—to give reasons for their conclusions even on credibility. The reasons must be "cogent" and "specific," Gailius v. INS, 147 F.3d 34, 46-47 (1st Cir. 1998), and must be "reasonably grounded in the record, viewed as a whole," Cordero-Trejo, 40 F.3d at 487. It is not our task to "invent explanations that may justify" the agency's conclusion. Dia v. Ashcroft, 353 F.3d 228, 260 (3d Cir.2003) (en banc).

The government may be arguing that the quality of the explanation provided is irrelevant because, under a 1996 amendment, "administrative findings of fact are conclusive unless any reasonable adjudicator would be compelled to conclude to the contrary," 8 U.S.C. § 1252(b)(4)(B).4 The quoted language reflects a gloss by the Supreme Court as to when a court may disagree with the fact-finder's weighing of evidence. See INS v. Elias-Zacarias, 502 U.S. 478, 483, 112 S. Ct. 812, 117 L. Ed. 2d 38 (1992). Neither we nor other courts treat this language as eliminating the conventional requirement that an administrative agency's explanation be rational.5

In this case, Hurtado could have been a rogue officer acting without any direction, or knowledge of a planned atrocity could have been (for obvious reasons) deliberately withheld from Castañeda and others not directly involved in carrying out the killings. Such possibilities do not prevent a fact-finder from disbelieving Castañeda's denial of knowledge, but they underscore the need for a rational explanation, rooted in the record, to support a finding that Castañeda should be disbelieved.

On the present facts, a well-reasoned explanation by the IJ and Board would likely require, first, specific findings with record support that Castañeda lied, or evaded answering or was significantly inconsistent in his responses to subsidiary questions; and second, an inference that because he lied on these subsidiary matters, his denial of advance or contemporaneous knowledge of a massacre was also false. There are other ways to discredit a witness, but the two-step sequence just described—falsus in uno, falsus in omnibus —is the most plausible method in this case.6

The BIA and IJ together offered five specific reasons—demeanor aside—for finding that Castañeda was not credible. We consider each of the five in turn and find each flawed to varying degrees—several badly so as to reasoning, record support, or both. We summarize our conclusions as to each immediately below but leave the details of each, together with necessary transcript quotations, for an addendum to this decision.

Two of these reasons (the first and the fourth), although weak, involve debatable interpretations of evidence, and we might defer as to these subsidiary findings if they stood alone. But as to the other three, the IJ's and Board's explanations do not make sense and appear at odds with the record, and these flaws badly undermine our confidence in the IJ's and Board's ultimate conclusion that Castañeda was not credible in denying knowledge.

The BIA's and the IJ's first reason for finding Castañeda not credible is that he failed to produce the transcript from his military tribunal proceeding and, according to the IJ, Castañeda was "extremely evasive and non-responsive" when questioned as to why. The IJ offered no specific examples of evasion; and careful inspection of the transcript suggests that Castañeda, who has no legal training, was simply ignorant as to the existence of a record and confused as to what was being asked of him.

Second, the BIA and IJ found that Castañeda gave inconsistent testimony concerning whether he was in radio contact with Hurtado's and Riveri Rondon's patrols during the massacre. The IJ's explanation, which the BIA summarized, is based importantly on the IJ's mischaracterization of the record and the attribution to Castañeda of statements that Castañeda did not make. The demeanor evidence, related only to this issue, was that Castañeda blinked more rapidly than usual.

Third, the IJ and BIA found that Castañeda was evasive about the extent of human rights abuses committed by the Peruvian military. In fact, the transcript shows that Castañeda was consistent in answering that, while he knew that some human rights violations had occurred, they were not the norm. This was not the answer the prosecutor sought, but it was not evasive, vague or inconsistent.

Fourth, the BIA and IJ found that Castañeda was evasive about the judgment issued against Hurtado, the commutation of Hurtado's sentence, and Hurtado's reappointment and promotion. Whether Castañeda's testimony was inconsistent or incomplete in some details is a judgment call; his basic position was consistent. Castañeda's lack of mastery as to detail was hardly surprising since Castañeda left Peru two years before Hurtado was sentenced, and four years before Hurtado was reinstated and promoted.

The IJ's fifth and final ground—not adopted by the Board—was that Castañeda was "incredible" in saying that he learned of the massacre only three weeks after it occurred, given that he was debriefed after he returned to base. The IJ's finding has no basis in the record: the massacre might well have been concealed from anyone not directly involved for as long as possible.

A State Department Human Rights Report, referred to by the agency, said that Hurtado and "three other officers" bore responsibility for the massacre. Castañeda was surely one of the three others. But the State Department report cited only the Senate Report and a preliminary military investigation not otherwise identified. Since the Senate Report (quoted above) made no finding as to Castañeda and the final military inquiry dismissed the charges against him, the State Department reference provides no basis for inculpating Castañeda and was not relied upon by either the IJ or the BIA as a basis for discrediting him.

We do not say that the evidentiary record "compels" the IJ and the Board to credit Castañeda's denial of prior or contemporaneous knowledge. Rather, in explaining their conclusion, the IJ and Board misstate Castañeda's testimony, apply labels (like inconsistent and evasive) that are at odds with what the transcript shows, and draw inferences that appear wholly speculative and without record support. A remand for further consideration is the proper remedy.

The suggestion may be made (see, e.g., He v. Ashcroft, 328 F.3d 593, 603-04 (9th Cir.2003)) that remand gives the agency a second bite at the apple. The short answer is that, outside criminal prosecutions governed by double jeopardy principles, second bites are routine in litigation. If the agency decision is flawed by mistaken legal premises, unsustainable subsidiary findings, or doubtful reasoning, remanding to give the agency an opportunity to cure the error is the ordinary course.7

By contrast, if the record compelled the IJ and the Board to believe Castañeda, it would be appropriate—although not necessarily required—for us to treat the issue of knowledge as definitively resolved in Castañeda's favor. We might then limit the remand solely to the issues that would remain (primarily, whether Castañeda otherwise qualified for asylum). Naturally, the barrier to such a determination is high—and it would be high even without the restrictive language that Congress has adopted.

Castañeda urges that great weight should be given to the Peruvian military court's dismissal of charges against him, and says that the BIA erred in characterizing the proceeding as a whitewash. Castañeda supplied no information about the circumstances of the acquittal or how the grounds or evidence related to the central issue here, namely, Castañeda's knowledge. Moreover, the government submitted evidence calling into question the impartiality of the tribunal, which the IJ and Board were entitled to credit.

The only remaining affirmative evidence for Castañeda is his own testimony. Although we think the IJ's and Board's parsing of his testimony is seriously flawed, this does not mean that a more cogent explanation would necessarily fail; nor is the Board or IJ precluded on remand from reopening the record and taking further evidence. 8 C.F.R. §§ 1003.2, 1003.23. In the end, the burden remains upon Castañeda to persuade the fact-finder to believe his denial of culpable knowledge.

One of the amici urges us to alter the deferential standard of review conventionally applied in these cases, saying that the immigration courts have a poor track record and should not be trusted. The framework for review is largely dictated by statute and binding Supreme Court precedent. E.g., 8 U.S.C. § 1252(b)(4); Orlando Ventura, 537 U.S. at 16, 123 S. Ct. 353. We will not raise the bar, but will insist that the agency clear the bar and think it has not yet done so here.

On remand, the IJ and Board are free to adopt the position, or to assume arguendo, that knowledge is required in this case and then to explain plausibly why it disbelieves Castañeda's denial. They are also free to adopt (and then seek to defend on appeal) a legal standard as to scienter different than the presumptive one that we have framed, or to take additional evidence or to do both. If they do alter the standard, they may have to provide a new evidentiary hearing so Castañeda can seek to meet it.

People are sometimes skeptical about whether an agency, asked by a reviewing court to revisit a matter, will do so with an open mind. We count on the agency to give this case a truly fresh look, unclouded by any prior judgment of Castañeda's credibility. Although we intend no criticism of the IJ, and the matter of assignments is primarily for the agency, confidence in this instance would be enhanced if a new IJ were to make the next assessment.

The decision of the IJ and the Board excluding Castañeda from asylum and/or withholding of removal under the persecutor bar is vacated and the matter is remanded for further proceedings consistent with this opinion.

It is so ordered.

In this case, the BIA and IJ together offered five reasons for finding Castañeda not credible:

First, that he was evasive when asked to explain why he failed to provide a transcript of his military court-martial;

Second, that he offered inconsistent and evasive testimony concerning his radio contact with Hurtado and Riveri Rondon's units during the Accomarca operation;

Third, that he was evasive concerning his knowledge of the extent of human rights violations by the Peruvian military;

Fourth, that he was not forthcoming concerning the judgment against Hurtado and Hurtado's subsequent reappointment to the military; and

Fifth, that he was not credible concerning the manner in which he claimed to have found out about the massacre.

In reviewing the transcript, there are some occasions in which it appears that answers or inferences from answers may have been affected by problems of translation and transcription. We have noted these in the discussion that follows as part of our attempt at an independent assessment. At the same time, it is fair to note that Castañeda has not complained about these on appeal and they are to some extent judgment calls concerning which the IJ—who heard the testimony—has the superior vantage.

1. Evasion Concerning Court-martial Transcript

The BIA and IJ's first reason for finding Castañeda not credible is that Castañeda did not produce the transcript from his military tribunal proceeding and, according to the IJ, he was "extremely evasive and non-responsive" when the prosecutor questioned him as to why he failed to do so. However, the IJ offered no specific examples of evasion, and the transcript suggests that Castañeda—who has no legal training—was confused as to what was being asked of him and ignorant of the existence of such a transcript, not evasive:

Q. Sir, are there any written documents or reports that confirm your version of your actions in the Accomarca incident?

A. Yes. I want to know regarding what documents about what?

Q. I'm asking you if there is any report or written accounting of the facts that support your version of your location that day, the actions by your patrol?

A. I have none of these documents. What I have is the decision of the martial court without the documents that I presented in that court remain there.

Q. The prosecutor appeals the decision of the court and the appeals court affirms what the trial judge did, correct?

A. Yes. The judge's verdict that we were innocent. We were exonerated

...

Q. The government appealed that decision, correct?

A. I did not know if the government appealed that decision or not.

Q. But the decision was appealed and the appeal court made a decision, correct?

A. I don't know very well what the procedures about that. I don't know if the government makes an appeal and it moves to a court of—a supreme military court or what I have was an interview with the human rights commission where I was considered a member of that massacre and with the previous investigation. Later on, I was taken to martial court and as I said on four or five occasions and after that, jury or judge, after that I was found innocent.

PROSECUTOR TO JUDGE: This is non-responsive. I asked about an appellate record.

JUDGE: Sir, answer the question. Do you know who appealed the decision?

A. I did not know who appealed.

PROSECUTOR: But there was an appeal.

A. In the document it manifests an appeal. . . .

Q. Do you know what documents, what record the appeal court looked at to decide what the facts were so they can decide that the dismissal was correct. . .?

A. What I know is that the commission of human rights, I don't know if they represent the government. I don't understand that very well. I understand it was in the process of appealing, appeal.

JUDGE: Sir, answer the question if you know the answer. Do you know what records the appeal court used, yes or no?

A. I do not know.

PROSECUTOR: The investigation that the army conducted was in writing, correct, was reduced to writing? Have you ever seen that document?

A. No.

Q. Did your defense attorney ever ask the government to produce that in your defense?

A. I don't know if he presented something.

Q. Did you ever ask for a copy of this investigation to assist in your defense at this court-martial? . . .

A. No. I did not request documents.

Q. Why not?

A. Because I did not know what sort of documents I need to ask or—

Q. No further questions . . .

2. Inconsistent Testimony Regarding Radio Communications

The BIA and IJ found that Castañeda gave inconsistent testimony concerning whether he was in radio contact with Hurtado and Riveri Rondon's patrols during the massacre. The IJ's explanation, which the BIA summarized, is not supported by substantial evidence in the record. The IJ reasoned as follows (with key passages not supported by the evidence underlined):

[T]his court finds that [Castañeda] provided inconsistent and evasive responses to his activities regarding radio communication during the massacre . . . .

During the first part of [his] cross-examination, he testified unequivocally that it was absolutely crucial for patrols during military operations to communicate with each other so that they did not have accidental contact . . . . However, [Castañeda then said] that he was not provided the radio frequencies for the other patrols on that operation at Accomarca. This is completely far fetched. . . .

[Castañeda] testified very clearly in his earlier testimony that coordinating an attack with patrols from different bases would require each patrol . . . responding to each other. Yet . . . [h]e wants this Court to believe that he was simply in his blocking position without communication with anyone. Yet, later in his cross-examination, he testifies that he extracted from his position upon communication [with] his commanders.

Early in his cross-examination, he testified that the patrols communicated with each other to avoid accidental contact. Yet, he would have this Court believe that not only was he not in radio control, but he was never given the frequencies that [the attacking patrols] were operating on. . . .

It is completely illogical that a blocking force during a military operation . . . to block an escape route, would not be informed by radio of the attack on the target and of the status of those being attacked, that is whether or not they were fleeing. . . .

When the Government did question [Castañeda] as to why he did not have the radio frequencies . . . [his] answer was completely non-responsive and evasive. He spoke of how the attacking patrols could talk to each other, but not the blocking patrols. He . . . never answered the Government's question.

In addition, the IJ noted that Castañeda's "demeanor during this part of his testimony" revealed him to be untrustworthy; specifically, the IJ noted that Castañeda was "blinking his eyes in an unusually rapid rate as compared with the rest of his testimony."

Much of the IJ's reasoning mischaracterizes the record. Castañeda never testified that "it was absolutely crucial" for patrols to communicate with each other. He said that it was "very important" to know the location of other patrols "to coordinate the movement," and that he would "keep contact with the base by radio" (emphasis added) to remain apprised of the location of other patrols. The government lawyer asked, "Would you also make direct contact via radio with other patrols in your area?" Castañeda replied, "Sometimes"— not always.

There is no inconsistency between this colloquy and the later exchange concerning the Accomarca operation, in which Castañeda stated that he had no direct radio communication with the attacking patrols. According to Castañeda, in the ordinary course, patrols would remain in contact with, and receive orders from, their base; they would not be in direct contact with other patrols. This is consistent with his not having been provided radio frequencies to communicate with the attacking patrols.

Contrary to the IJ's assertion, Castañeda never testified "very clearly . . . that coordinating an attack with patrols from different bases would require each patrol . . . responding to each other." Rather, when asked how he coordinated an attack with other patrols, he explained that communication between patrols was indirect, traveling up and then back down the chain of command:

Each base has a patrol. The patrol will go and conduct an operation. Each head of a patrol will respond to the immediate boss, the head of base. The head of base at the attack, they will inform the command of the battalion who can communicate and coordinate directly with the head of patrol and give another specific order or can get contact among the heads of patrol.

The government's counsel pressed, expressing skepticism that patrols would be unable to communicate with each other. Castañeda's response was not pellucid— due at least in part to what appears to be poor translation or transcription—but the gist again appears to be that patrols report to the heads of base:

There are different situations. . . . [T]he organization of battalion against the guerrillas is completely different [than in a conventional war] and the company is segregated in base in an expansion in an area. So the firing of corroboration, they are through radio in coordination with heads of base and also transmitted to their patrol. One time, a patrol was moved from a base by helicopter to the other base to coordinate and attack a mission together, but sometimes depending where the mission was, there was just the base . . . . Each base had a [topographical chart] and they will indicate the point for the patrol where they are at and for a coordinated force.

Castañeda also never testified that "he was simply in his blocking position without communication with anyone." Rather, he consistently testified that he was in communication with his base, but not with the attacking patrols. The colloquy was as follows:

Q. How were you told [that the attack had ended]?

A. They called me by radio and the captain told me I could go back.

Q. Were you monitoring the radio?

A. I was listening to my radio operator.

Q. Were you able to pick up radio traffic between the two teams that were in the attack?

A. No. I was not able to hear their communication . . . .

Q. Did you hear the communication when the platoons called their headquarters to . . . let the headquarters know they were ready to [begin] the attack?

A. No.

Q. You and the radio operator heard no other radio traffic.

A. No. I did not hear any radio attack. . . .

Q. Who decided what frequency your radio would be on for communicating with base?

A. There is a frequency established by base. . . .

Q. Were you were [sic] what frequency you needed to communicate with these other patrols in the case of an emergency, correct?

A. To communicate with my base.

Q. For example, if one of those patrols needed to come towards your position, it would be helpful if the two of you knew how to communicate, correct?

A. That's true.

Q. So you were informed what frequencies the two attacking patrols would be using, correct?

A. That's correct, but both patrols entered from the same place and where those patrols are directed, four patrols are directed by the same head of unit, but these orders came from the division. I had directly directed by the head of commander of base.

Q. But who's got operational command of this attack?

A. The head of the division. . . .

Q. Wasn't there a division frequency that they could call you on?

A. I had only one frequency with my base.

JUDGE. Sir, it still hasn't been answered though. Could you contact the other two patrols by radio, yes or no?

A. No, I did not on my radio.

JUDGE. That's not what I asked. Could you contact them by radio, yes or no?

A. No. I could not.

JUDGE. Why not?

A. Because I did not have their frequency.

JUDGE. Why not?

A. Because they didn't give them to me. I only used to communicate with the head of my base that was higher. . . .

Q. Did the attacking patrols know your frequency? . . .

A. No.

Q. How do you know that they did not have your frequency?

A. Because they never relayed anything. They only responded to their head.

Finally, again, taking account of what appears to be poor translation and transcription, Castañeda was not "completely non-responsive and evasive" in answering the government's query as to why he lacked the radio frequencies of the attacking patrols. The government's counsel, in a long question, asked Castañeda to reconcile the importance of coordination with the fact that he lacked radio frequencies for the attacking patrols. The question was followed by an objection and back-and-forth between Castañeda's counsel and the IJ. Finally, Castañeda gave an answer that emphasized advance planning as the keystone of coordination rather than inter-unit communication, but the answer was clearly infected by poor translation:

A. The planning is very important for communication. At the time, there was a planning that the commander from the other base the two patrols were closely—to the point where they need to close those point and that they should not move from there, just need to close those point. There are two patrols in—

INTERPRETER. Your Honor, the Interpreter will need to ask him about that.

JUDGE. Go ahead.

CASTAÑEDA. To patrol, to transport so they can act as with direct orders from the division their independence from the order that was given to the heads of both bases so to cause as few possible exit points and they planned it that.

Q. Sir, how do you exercise judgment independent of the bases when you can't communicate with any of the other units attacking the objective? Are you sure none of those patrols could talk to each other?

A. Yes. The attacking, the patrols could communicate with each other. They were going to talk to the—the base. I have the same frequency at my base, but correspond with the sam[e] battalion.

As for the IJ's assertion that Castañeda blinked during his testimony, even if this demeanor observation is given some weight, the number of flaws in the IJ's reasoning and characterization of the record is enough for us to find his explanation wanting.

The government in its brief says that even if Castañeda's testimony was consistent, it was illogical: because the Accomarca operation was coordinated by the division rather than the battalion, and because it was one of Castañeda's first combat patrols, radio contact among patrols would seem especially important. But this explanation was provided by neither the BIA or the IJ (aside from the IJ's bare assertion, without any analysis, that Castañeda's account was "far-fetched"). Nor does it find any support in the record: given Castañeda's consistent testimony that radio communication ordinarily took place up and down the chain of command rather than among patrols, inter-patrol contact would seem even less likely than usual in an operation where patrols were drawn from separate battalions.

3. Castañeda's Knowledge of Human Rights Abuses Generally

The IJ and BIA found that Castañeda was evasive about the extent of human rights abuses committed by the Peruvian military in the emergency zone. The IJ stated, "At one point at the beginning of the Government's cross-examination on this issue, [Castañeda] testified that he is aware and was aware that human rights violations were occurring. . . . However, moments later, [he] changes his testimony, becomes significantly vague in his answers, and in fact disagrees with the State Department['s description of] hundreds upon hundreds of killings. . . ."

In fact, inspection of the transcript reveals that Castañeda was consistent in his answer: he knew that some human rights violations occurred, but they were not the norm. While this answer did not dovetail with the knowledge that the prosecutor wished to impute to Castañeda, neither was it evasive, vague, or inconsistent. Several times in the questioning the IJ interjected, asserting that Castañeda had not answered the question posed to him; but the transcript shows that Castañeda did in fact answer the question posed. The colloquy follows:

Q. Prior to August 1985, did you become aware of the security forces meaning the Peruvian army engaging in extrajudicial killings, disappearances, torture . . . and arbitrary detentions? Are you aware of that?

A. Yes.

. . .

Q. Most of these violations occurred in the emergency zone. Is that correct?

A. Some of it. It was not the norm.

JUDGE. Sir, I don't understand that answer. The question was, did most of these violations that you are aware of like the extrajudicial killings and torture by the Peruvian army occur in the emergency zone?

A. Could you let me explain or do you want just the plan [sic] answer?

PROSECUTOR. Can you answer that yes or no?

A. I did not know that. I know of very rare cases . . . .

Q. Did you discuss these cases with your commanders or did your commanders discuss them with you?

A. About what?

Q. The Peruvian army engaging in extrajudicial killings and torture of Shining Path guerrillas.

A. The information of the torture, we didn't unless it was some military. They inform us all this. That's what I knew was the minority of some elements outside of the rule, a minority. They have some of it, but not the entire Peruvian military. Sometimes, by some bad elements that belonged to this organization and they have been retired military.

JUDGE. Sir, that does not answer the question. . . .

PROSECUTOR. I'll move on, Judge. . . . [Citing the State Department Human Rights Report:] Total deaths related to any terrorist operations of the security forces are estimated by human rights groups as being in the thousands from 1980 to '85. Does that sound incorrect?

A. Correct.

Q. What were you aware of prior to August 1985?

A. I knew that some very few bad element.

Q. Sir, could I have specifics? What specifically did you know about army personnel committing human rights abuses prior to August 15, 1985? What were you aware of? What had you heard? A. That there were a number of officers that had detained terrorists or possibly terrorists in an unjust manner.

Q. What about killings, sir?

A. Officers that had detained prisoners and that they were executed.

Q. What about civilians suspected of supporting [the Shining Path]? . . .

A. . . . I did not know about detained civilians. I did not know that they had detained civilians accused of helping and then being killed.

Q. What about suspected [Shining Path] terrorists being executed or tortured?

A. I'm talking about some official. I heard of that acted on—acting with no orders, taking justice in front of the military group.

Q. Sir, you were aware that this kind of thing happened on occasion prior to August 15, 1985.

A. Occasionally, not constantly.

Q. Were you aware of any incidents that happened in your area of operation in anti terrorist battalion 34? . . .

A. In others. I participated in this operation.

Q. I'm talking about prior to this operation.

A. No. I did not know.

Q. We're talking about just in the area of battalion 34.

A. I did not know.

The BIA and IJ found that Castañeda was evasive about the judgment issued against Hurtado, the commutation of Hurtado's sentence, and the reappointment of Hurtado to the military. The IJ reasoned that Castañeda

in a very evasive way . . . stated that . . . anyone who is convicted of such a crime like Lieutenant Hurtado would be automatically dismissed from the military. Yet, when [Castañeda] was presented moments later by the Government with evidence that Lieutenant Hurtado was in fact never convicted of murder, but was only convicted of abuse of authority, [Castañeda] then immediately changed his testimony and admitted that he had heard about that through the media. . . .

However, when given the opportunity by the Government prosecutor moments earlier to tell the Court what he knew about Lieutenant Hurtado, [Castañeda] did not do so. This goes directly to his credibility.

In fact, Castañeda did not change his testimony. He testified that he believed Hurtado had been convicted of murder. When presented with a State Department Human Rights Report indicating that Hurtado had in fact only been convicted of abuse of authority, Castañeda responded: "Clearly, I don't know. What I manifested what I know is that [the] court found him guilty of that murder and I'm not an expert on the law, but maybe the murder is instead of the disobedience, sorry, abuse of authority."

The prosecutor also asked Castañeda whether Hurtado had been discharged. Castañeda responded, "I do not know really if he was discharged. What I know is that he was recluded in prison." After some additional questioning, Castañeda added, "[E]very officer expected for a situation like in this case, he's automatically discharged. That's what I know. I did not try to find out what privileges he might have had and what had happened to his life." A little later, the prosecutor quoted from a State Department Human Rights Report to the effect that Hurtado had "reappeared on active duty," and the following colloquy took place:

Q. Do you know whether Lieutenant Hurtado was promoted to captain?

A. I did not know that. I find out through public media when I was here in the United States, but he was not involved in activity which I did not know was true or if he was absolved by some law. I did not know that.

Q. Would you surprise you to know that he was promoted after been convicted and sentenced?

A. Yes. It was a surprise.

Later on, when asked again about his knowledge of Hurtado's release from prison and promotion, Castañeda further clarified:

A. I did not know he was promoted to captain. I think he was not absolved. I only heard the comments in the news and the TV.

Q. What comments did you hear on TV?

A. I heard the comments that he was promoted to captain and that the army was demanding that.

Taking his testimony as a whole, Castañeda testified consistently that he did not know whether Hurtado was actually discharged, even though that would be the normal procedure in a case such as his; and that he was surprised when he learned from the news media that—in contravention of ordinary procedures—Hurtado had apparently been promoted. Nor is it surprising that Castañeda knew few of the details concerning Hurtado's sentence and ultimate reinstatement and promotion. Hurtado was sentenced in 1993, two years after Castañeda left Peru, and was not released from prison and reinstated until a general amnesty in 1995.

The testimony was not flawless: one could perhaps infer that Castañeda had held back his knowledge of Hurtado's reinstatement until pressed on the subject by the government. Whether one regards this as significant, and infers from it that Castañeda was evasive, is very much a judgment call.

5. The Manner in which Castañeda Learned of the Massacre

The IJ's final ground for finding Castañeda not credible concerned Castañeda's testimony about how he learned of the massacre: he testified that he first learned of the killings three weeks later on the radio. The IJ found it "incredible that although [Castañeda] reported back to his base and although he was debriefed, he was never aware that approximately 69 civilians were raped and murdered by the other two patrols that he claims were four or five miles away."

This is pure speculation by the IJ: there is no basis in the record for the IJ's conclusion that because Castañeda was debriefed following the operation, he would necessarily or even likely have been informed about a massacre committed by the head of another patrol. The Board chose not to rely on this ground in affirming the IJ's general finding that Castañeda was not credible.