The defendants were convicted by a jury of armed robbery of a bank under 18 U.S.C. 2113. On appeal we affirmed. 269 F.2d 688 (1959). The Supreme Court granted certiorari, but limited its decision to the single issue of whether the district court had properly complied with the Jencks statute, so-called, 18 U.S.C. 3500, in refusing the defendants access to an F.B.I. interview report of one of the government's witnesses, one Staula, for the purpose of impeachment. The court, in effect, suspended the convictions and returned the case to the district court for the purpose of making a further factual inquiry into the circumstances attending the report in order to determine whether defendants had improperly been denied access to it, or to any other paper. For the aid of the district court the court recited certain legal principles, but it left undecided what consequence might flow from non-production, because of prior destruction, of an otherwise properly called-for document. The court held, inter alia, that it had been error to refuse access to the report on the basis of Staula's denial of its accuracy because that permitted the witness to immunize himself from impeachment; instead it directed the court to conduct, in the absence of the jury, a 'non-adversary' hearing, at which Toomey, the F.B.I. investigator who prepared the report, would be cross-examined by the defendants. It held that it had been error for the court to tell the defendants that they could have Toomey only as their own witness. The court gave to the district court the option of calling Toomey itself, or ordering the government to call him. Campbell v. United States, 1961, 365 U.S. 85, 81 S. Ct. 421, 5 L. Ed. 2d 428.

The hearing before the district court held pursuant to this mandate occupied an extraordinary amount of time. This was largely because of needless remarks by the court, although we are not unmindful that counsel contributed. While technically the court called Toomey itself and permitted the defendants to cross-examine, the restrictions imposed upon counsel were such that it was cross-examination in name only.1 In spite of the fact that the witness was a special agent of long standing who had discussed his testimony with the Assistant U.S. Attorney immediately before the hearing, the court hovered constantly over him like an over-anxious mother. With respect to correlation between the notes, Staula's statements, and the eventual report, the Supreme Court's directions for a non-adversary proceeding to assist the court in performing its duty, with the defendants permitted to cross-examine, were honored largely in the breach. The necessary cure for this will be for us to make our own interpretation of Toomey's testimony so as to give the defendants the benefit of every doubt and to draw all proper inferences in their favor. This we will proceed to do.2

The robbery occurred on July 18, 1957. Staula was a customer present in the bank, and an eyewitness. Toomey interviewed him in a room in the local police station the following morning. No one else was present. They ast at a table. Toomey talked with Staula, asked him questions, and took notes. The notes did not purport to be word for word, or even to use Staula's own words, except when specifically indicated by quotes. Toomey did not use shorthand, but occasionally used symbols and abbreviations. He recalled two omissions-- the fact that when Staula put some money of his own that he had been about to deposit in his pocket he did so to 'out-cute' the robbers, and that he feared being locked in the vault. But in Toomey's opinion, the notes were 'complete * * * with respect to the pertinent information' Staula had given him. However, they were complete only with respect to subject matter; the notes were not a running account. This was further demonstrated by the fact that after Staula had finished telling his story and had answered Toomey's questions, Toomey recited back to Staula the substance of what Staula had told him. He did not do so by reading his notes. Though he relied 'primarily' on them, he depended also upon his memory. Staula answered in the affirmative Toomey's question whether he had the story straight. Toomey did not ask Staula to sign anything, or to initial anything, nor did Staula himself read the notes, although he was in a position at the table to have observed him writing.3

Toomey testified that the interview had taken place in the forenoon; that it lasted half an hour; that he attended to some other matters during the day, and that that night he dictated his interview report into a machine. He stated that he first arranged the notes in chronological order, and then relying primarily on his notes, but also on his memory, dictated a report that 'reflects the information in the notes.' However, he used his own language. He did not believe, for instance, that the phrase in the report that 'he (Staula) went to the teller's window which is served by Mr. Kennedy,' and the phrase, twice repeated, that he 'observed these individuals no further,' were the witness'.4 He testified that he sent the machine disc to the Boston office for typing; that he received the transcribed report back some days later; that he checked it against his notes, and on finding it accurate, destroyed the notes in accordance with standard F.B.I. practice. He did not show this report to Staula, and, in fact, never saw him again after July 19.

On this record it is entirely clear that Staula agreed to the accuracy (and 'approved,' to use the statutory language) of what Toomey recited to him as his understanding as to his, Staula's, account on July 19. But it seems also clear on Toomey's testimony that what Toomey so told Staula that morning differed to some extent from what was in his notes, and, on both Toomey's and Staula's testimony (and there could be no other, so far as appears), differed also from what he put in the eventual report, although all may have been summaries of the same thing. Before considering the importance or consequence of these differences, we review the applicable principles as we understand them.

Neither the Supreme Court in its opinion, nor the defendant presently, suggest that any paper here sought5 is producible other than in accordance with 18 U.S.C. 3500. Palermo v. United States, 1959, 360 U.S. 343, 360, 79 S. Ct. 1217, 1223, 3 L. Ed. 2d 1287. A statement is not producible under this statute unless it is,

'(e) * * *

'(1) a written statement made by said witness and signed or otherwise adopted or approved by him; or

'(2) a stenographic, mechanical, electrical, or other recording, or a transcription thereof, which is a substantially verbatim recital of an oral statement made by said witness to an agent of the Government and recorded contemporaneously with the making of such oral statement.'

This statute is by no means clear. It might well have been possible, as contended by the dissenters in Campbell, 365 U.S. 858 at 104, 81 S. Ct. 421, at 431, 5 L. Ed. 2d 428, to interpret subsection (1) as limited to a statement 'written * * * by said witness,' and subsection (2), in complementary fashion, as a statement given orally but written or otherwise transcribed by someone else. However, the court has construed subsection (1) to include a statement orally made by the witness and written by someone else. This broader construction must, in turn, cast light upon subsection (2). We must hold this latter to be limited to oral statements by the witness taken down contemporaneously by someone else in a particular fashion which produces a 'substantially verbatim recital.' Thus subsection (1) applies if the writing, even though made by someone else, was 'signed or otherwise adopted or approved,' although it did not follow the original statement substantially verbatim, while subsection (2) applies if the writing was substantially verbatim, even though it was not signed, adopted, or approved. This, in turn, must mean that a writing not specifically approved is not producible in spite of its general accuracy if it is not substantially verbatim.

Before construing the particular statutory provisions it may be appropriate to turn to the legislative history. While it has sometimes been said that the only purpose of the act was to rectify lower court misinterpretations and enlargements of the Jencks decision, we must observe that public clamor about the consequences of that decision, and the incipience of Congressional action, was immediate and was, at best, only prophetic so far as judicial misunderstanding was concerned. See 103 Cong.Rec. 8290 (June 4, 1957). Eleven bills were introduced. It is evident that the primary consideration was protection of the F.B.I. A review of the Congressional history suggests, as is often the case in important legislation, that views differed as to the extent of the legislative protection desired. We believe it apparent that some of the proponents were more concerned with protection than they were with merely 'interpretation,' and that their principal deterrent was apprehension as to the scope of due process.

Examining the statute as a whole, it might be argued at first blush that the only aim was the elimination of confidential matter and irrelevancies, as provided by section (c). However, it also appears that not every memorandum made by an F.B.I. agent as the result of an interview is to be indiscriminately reached even though it contains no such material. The F.B.I. has a legitimate interest in maintaining secrecy on a broad base. But it also may have an interet to some degree in the purely mechanical aspects of its operations in permitting its agents to take notes in such manner as they see fit (not to mention destroy them), or to make memoranda for their own use, without the spectre hanging over them that such papers, with all of their possible imperfections, may be called for in court. Vivid, also, in the minds of all trial lawyers are the related problems of work products, Hickman v. Taylor, 1947, 329 U.S. 495, 67 S. Ct. 385, 91 L. Ed. 451. It can be readily understood that if a memorandum of a conference prepared by an F.B.I. agent which did not attempt to be substantially verbatim, and that had not been specifically adopted by the interviewed witness, is to be subject to production in court, freedom of operation of the agents would be considerably curtailed. The framers of the bill may well not have wanted that impediment.6

The defendants complain that any restrictive interpretation would destroy the purposes of the act, which they assert was designed to protect prospective court defendants. It was even intimated that it was not only the duty of the F.B.I. under the act to preserve notes, but it was their duty always to take notes, so that a record might be there to be kept. We could not entertain such a suggestion seriously. By the same token, we cannot believe that if the F.B.I. does take notes or make reports it must do so in such accurate form that they will be safely producible in court, or suffer the consequences. If, for any reason, valid impeachment material does exist, a defendant is entitled to have the benefit of it. But this is, in a sense, a windfall rather than the performance of a duty owed. The word 'windfull' may sound alarming, but either there is a duty owed, or there is not. We do not find in the statute any such duty. The statute speaks simply of statements, meeting certain specifications, 'in the possession' of the government, not statements which should be, or should have been.7 The placing of new affirmative duties upon the F.B.I. would be entering a field for which we find no intimation in the legislative history, or justification in the statute.8

Nor can the unavailability to defendants of statements which the witness did not specifically approve, or which were not substantially verbatim, represent a serious miscarriage of justice. Just as soon as a question arises as to whether a writing does in fact represent accurately the prior statement of the witness, collateral issues arise. If, in point of fact, the statement was not approved by the witness, and is not a completely accurate representation of what he said, the resulting 'impeachment' may be highly unfair.

Turning to the specific provisions, we do not find the words 'substantially vertatim' easily construed loosely as, for example, 'substantially accurate.' We are not sure how far the reverse of the negative expressed in Palermo v. United States, 1959, 360 U.S. 343, 352-353, 79 S. Ct. 1217, 3 L. Ed. 2d 1287, carries. Except to the extent that we must do so, we would not be disposed to agree with the dictum in United States v. Thomas, 2 Cir., 1960, 282 F.2d 191, 194, that a report prepared as a 'substantially accurate reproduction of agent's original notes' may be 'substantially verbatim' within subsection (2), particularly if that includes, as the court said in United States v. Waldman, D.C.N.J. 1958, 159 F. Supp. 747, at 749, (cited with apparent approval in United States v. McKeever, 2 Cir., 1959,271 F.2d 669, 674) 'an elaboration' of the notes. See also United States v. Annunziato, 2 Cir., 1961, 293 F.2d 373, 381. We believe that subsection (2) is directed to statements, which, lacking approval, substitute the guarantee of some means of transcription ('stenographic, mechanical, electrical, or other recording') which lends itself to a very high degree of exactness. This, in turn, permits a definite, and limited, inquiry by the trial judge into the efficacy of the means of transcription, rather than the indefinite, and often unsatisfactory pursuit of whether the writing was, in over-all substance, accurate. Congress used very precise language. ('Detailed particularity.' Palermo v. United States, supra, 360 U.S. at 349, 79 S. Ct. at 1223.) If all it meant was 'substantially accurate,' it would have been very much easier to say so. To us the length of debate on the subject of records vs. recordings, supra n. 6, is very revealing. A longhand writing which the court found fairly followed the witness' words, subject to minor, inconsequential errors, as is any 'stenographic' reproduction, would fall within subsection (2). Neither in his notes, nor in his interview report, did Toomey even purport to do this.

The requirements of subsection (1) are quite different. Here Congress required approval (of non-substantially verbatim statements) by the witness' signature, or 'otherwise.' This is not directed to the agent's substantial accuracy, but to the action of the witness. In the case at bar, always considering Toomey's testimony by itself, it is true that Toomey wrote something before parting company with the witness. But he did not purport to show him the words that he had written, or to read those words to him. Lawyers and trial judges are quite familiar with witnesses who are willing to read over a written statement and state orally that it is correct, but are unwilling to affix their signature. But this is not the same as having an incomplete writing and elaborating it in recounting back to the witness what he is understood to have said, or then subsequently reducing the interview to writing in undoubtedly not the same words. Where the statute says 'signed, or otherwise adopted or approved,' by ordinary principles of noscitur a sociis, we think this means an approval comparable to a signature, and refers to the written statement itself, not merely approval of a general account of which the writing may be representative.

Again, if Congress meant less than this, there would have been no need of these two elaborate subsections. All that would have been required would have been a single phrase, 'substantially accurate reproduction or recording of what the witness said.' It could even have been left to the courts to supply the obvious, that in cases where the witness had in fact indicated his approval, this was evidence enough. We believe that Congress has imposed far more definite safeguards.9 /9/ See United States v. Thomas, supra, 282 F.2d at 194, as to subsection (1).10 On subsection (2) our decision is in accord with Borges v. United States, 1959, 106 U.S.App.D.C. 139, 270 F.2d 332, cert. den. 361 U.S. 971, 80 S. Ct. 601, 4 L. Ed. 2d 550, and United States v. Stromberg, 2 Cir., 1959, 268 F.2d 256, cert. den. 361 U.S. 863, 80 S. Ct. 123, 4 L. Ed. 2d 102. See also the somewhat ambivalent decision of Papworth v. United States, 5 Cir., 1958, 256 F.2d 125, cert. den. 358 U.S. 854, 79 S. Ct. 85, 3 L. Ed. 2d 88, which contains a dictum that notes of 'the highlights of the conversation' are 'substantially verbatim,' but holds that 'a running summary * * * in the language of the agent' is not.

We turn to the absence, over defendants' objection, of Staula at the hearing. This may have had some pertinence on the issues relating to the notes. It is clear on Staula's own testimony at the original trial that he never saw, and hence could not under our definition have approved, the investigative report as such. Similarly, with respect to the report, his testimony did not help defendants on the substantially verbatim issue. ('There are things in there turned around.' 365 U.S. at 97, 81 S.Ct. at 427-- a seeming protest, were it admissible, against the consequences of Toomey's rearrangement of his notes in chronological order.) However, the Supreme Court has indicated, without deciding, that if the notes themselves satisfied the requirement of the statute, the government may have to face some penalty for having destroyed them. The court did not wish to rule on this question on an empty record, and requested further facts.

Sufficient facts now adequately appear for us to decide that the notes were not substantially verbatim. On this issue Staula, the only other suggested witness, could not cast any further light. But although his testimony at the original trial was weak, on further examination he might have satisfied a trier of facts that he had read Toomey's notes and signed or otherwise approved them. The defendants had a right to pursue that matter at the hearing just as much as they had a right to cross-examine Toomey. They asked the court to summon Staula, but the court replied that the Supreme Court had said this could not be done. This was a complete misunderstanding of the opinion. It was on an entirely different issue that the court had made such a ruling. At the same time the district court informed counsel that it would consider Staula's testimony already of record, an incomprehensible inconsistency on its part. Thereafter it disbelieved that testimony. Although it had a right to disbelieve it, the defendants had a corresponding right to have Staula 'live' and cross-examine him further. While in our view the issue might be of no legal consequence, we must remand the case for further hearing and findings, with Toomey and Staula both to testify, as to whether Staula signed or otherwise adopted or approved the notes, in order that the mandate of the Supreme Court be fully complied with. Because of the committed position of the present district judge on this question, that hearing should be before another judge.

While retaining jurisdiction of this appeal generally, an order will be entered returning the original papers to the District Court with direction to have further proceedings in accordance with this opinion and to return said original papers including the further proceedings when they shall have been determined.