Evidence Without Custody

Three weeks before Jason Smith’s trial, the prosecution changed its expert witness. Elizabeth Rast had examined the seized substances at an Arizona state laboratory, recording her work in notes and a report. She had since left the laboratory. Greggory Longoni, another forensic scientist there, would testify instead. The State told the court that he would give an independent opinion and was “expected to have the same conclusion.”¹

Longoni knew chemistry and the laboratory’s procedures. Rast’s notes gave him considerably more to examine than a verdict written on a sheet of paper. They described the items, identified the methods used, recorded results and drew conclusions. For some tests they recorded a blank, a check for contamination. She had tested eight items and prepared a signed report distilling her findings. A colleague could read that work, understand what the procedures were intended to establish and consider whether the conclusions followed. Keeping such records is one of the ways a laboratory remains useful after a particular scientist has gone home, changed jobs or died.

At trial, Longoni gave opinions about four of the items. The questioning about Item 26 moved through the laboratory’s practices to the work Rast had recorded. Were the usual policies followed? Yes. What method had been used? Microscopic examination and a chemical color test. Had that been done in this case? “Yes, it was.” Had a blank been run? “According to the notes, yes.” Asked to bring his training, experience and review of the records to bear, he identified the item as a usable quantity of marijuana.²

The small phrase according to the notes belongs to the difficulty. Longoni had something to say about those notes. He had not performed the tests or observed Rast performing them. His understanding of what a test could establish and his account of whether it had occurred came by different routes, although the examination invited the jury to receive both through the authority of the same witness. He could be questioned about his chemistry. A question about what Rast had actually done reached a limit in his personal knowledge.

It is easy to mistake that limit for a defect in recordkeeping. Yet the record was precisely what made this testimony possible. More elaborate notes could have exposed a mistake, preserved a result for another interpretation or shown that a necessary step had been omitted. They could also have recorded, accurately, work that Longoni would never be able to recall doing. The distinction persists even when everyone has behaved competently. A report enables another person to know more than they witnessed; it does not give them every kind of knowledge the witness might have supplied.

Smith’s objection reached the Supreme Court as a question about the use of those earlier statements. Arizona’s courts had treated the information from Rast’s records as explaining the basis of Longoni’s opinion, rather than as evidence offered to prove that what she reported was true. That distinction would keep it outside the relevant hearsay inquiry. Justice Kagan’s opinion rejected the maneuver: where the absent analyst’s statement supports the expert’s opinion only if it is true, calling it the basis of the opinion does not change the work it is doing. The opinion could identify the material because Rast had reported certain tests and results. The jury was being asked to believe those things had happened.³

The Court’s June 2024 decision set aside the appellate judgment and sent the case back. It did not find that the tests were false or establish Smith’s innocence. A further constitutional question remained: were the particular statements testimonial, within the meaning of the Confrontation Clause? The Court left that question, and a dispute over whether Arizona had forfeited it, for the state court. Even which parts of the records Longoni had conveyed—notes alone, or notes and final report—was contested. A document kept for internal quality control and a statement made primarily for a criminal proceeding need not occupy the same legal position. The justices disagreed about how that position should be determined.⁴

What they confronted was therefore more exact than a choice between trusting a laboratory and distrusting one. The State possessed results it wanted to use, records of the work that had produced them and a qualified scientist willing to offer an opinion. It still had to establish what that scientist’s presence allowed the jury to hear.

An expert’s borrowed facts

Justice Alito, joined by Chief Justice Roberts, thought the majority had inflicted a needless injury on modern evidence law. Expertise is largely an organized dependence on other people. A scientist does not personally repeat the observations on which a discipline rests before using them. An expert’s usefulness can consist in knowing which findings deserve reliance, how they fit together and where an apparent inference fails. An insistence on firsthand knowledge at every step would disqualify much of what makes the expert worth hearing.

Courts had once commonly required the expert to answer a hypothetical question: assume these facts, then say what follows. The form advertised the distinction between a premise and its proof. In practice it also furnished lawyers with an opportunity to pack an argument into a question. Assumptions accumulated; the witness had to retain and interpret them; opposing counsel could object to their selection before an answer was ever given. Alito recalled a case in which a hypothetical occupied eighty-three pages of transcript and an objection another fourteen. A device meant to isolate the expert’s contribution could make it very difficult to discover what that contribution was.⁵

Modern rules made a more candid accommodation with professional practice. Under the federal comparison Alito used, an expert can rely on facts or data of a kind reasonably relied upon in the field even when those materials would not themselves be admissible. Disclosure of otherwise inadmissible material to the jury faces an additional test: its value in helping the jury evaluate the opinion must substantially outweigh its prejudicial effect. A limiting instruction, when requested, directs the jury to the permitted use. Cross-examination can expose that the expert lacks personal knowledge of an underlying fact. Alito defended these arrangements as ways of judging reliance, without pretending reliance had disappeared.⁶

Keeping the basis of an opinion out of view can make the opinion harder to assess. The listener needs to know what was borrowed to judge what the expert contributed. Alito believed a jury could use that disclosure to evaluate an opinion without treating each disclosed assertion as independently proved. The majority answered that, in the kind of testimony before it, the asserted facts supplied the very reason to credit the conclusion. An instruction about purpose could not sever that dependence.

But Alito and Roberts agreed that Longoni’s testimony had crossed a line. He had not confined himself to explaining an opinion formed from the records. He had affirmed that Rast followed particular procedures and performed particular tests. Neither his training nor his review gave him personal knowledge of those occurrences. Alito saw no reliable scientific method by which Longoni could infer their truth simply from her saying they happened. He would have disposed of the case on that narrower ground, subject to the same unresolved question about testimonial status.⁷

The disagreement leaves useful work for a second expert. Such a person may discover an inconsistency the first analyst missed, explain a result the first misunderstood or show that a stated conclusion does not follow even if every recorded observation is accepted. Longoni could also testify from his own knowledge about his laboratory’s ordinary practices. The majority expressly preserved that role. None of this requires pretending that an independent assessment is a second performance of the original experiment. Its independence lies in the work actually done: a judgment exercised on materials received, with whatever dependence on those materials the judgment retains.

That is already a considerable saving. It allows knowledge to be divided among people without requiring each of them to possess the whole. The trouble begins when an institution takes the saving once more, dispensing with a kind of support it still needs because the person presenting the conclusion has supplied a different kind well.

Let Cobham be here

The older demand to hear an accuser was directed at an institution that had made its own arrangements for knowing. At Raleigh’s treason trial in 1603, Cobham’s accusations reached the jury through an examination and a letter. Raleigh disputed not only their contents but the conditions under which the accusations had been obtained and the interests of the man who made them. In Jardine’s published account, he argues that Cobham stood in the king’s mercy: excusing Raleigh would do Cobham no good, whereas accusing him might earn favor. This was Raleigh’s proposed explanation, not a motive the surviving record entitles us to see inside Cobham’s mind.⁸

Raleigh wanted to question the man. When Lord Henry Howard told him that the request could not be granted and asked him to stop pressing it, the answer in Jardine is abrupt: “Nay, my Lord, it toucheth my life, which I value at as high a rate as your Lordship does yours.” The account then has Cecil ask whether further time might be taken to learn the king’s pleasure. The judges resolved to proceed. There was an accusation to be tried, an order of proceeding to be maintained and a life whose owner did not regard further inquiry as an inconvenience.

We receive this exchange through other people’s work too. Jardine’s 1832 edition identifies the reports from which it compiles the trial, prefers some to others and acknowledges gaps. In a nearby note he prints correspondence while admitting uncertainty about whether it was actually produced against Raleigh. The qualification does not make the edition worthless. It lets the reader distinguish a document that illuminates the affair from an event established to have occurred in the courtroom. The archive has not stopped us from learning. Its editor has told us where one inference must stop.⁹

When Crawford v. Washington returned to Raleigh in 2004, it put the emphasis on the refusal to permit confrontation. The prosecution had offered reasons for trusting Cobham’s statements; the constitutional objection, in Scalia’s account, was that the judges’ appraisal could not replace the procedure the accused was entitled to invoke. The Sixth Amendment, ratified in 1791, was not merely a recommendation that officials weigh credibility carefully. For the testimonial statements within its reach, it constrained the means by which the prosecution could make its case.¹⁰

This is a different claim from saying that a statement becomes true when its author can be questioned. A witness may lie under oath, remember badly or answer a good question badly. Nor does a refused confrontation prove the accusation false. The right matters because the government cannot make its own confidence in the accusation sufficient to extinguish the accused’s opportunity to test it. What the official is persuaded of and what the official is permitted to establish in that manner remain distinct.

It would be a mistake to turn this political history into a general hostility to secondhand knowledge. Raleigh was not asking every reader of his trial to reassemble the court at Winchester. And the law’s concern with absent speakers has always had to encounter people whose claims would never be heard if only the best form of evidence were admitted.

The cost of the better witness

Bentham gives that difficulty an inconveniently clear formulation. In his treatment of casually written evidence, he first recommends excluding the document where its author can be brought forward and questioned. His reason is that no information need be lost: the more searching encounter can yield what the writing contains and more. Then comes the exception that changes the character of the rule. If questioning is impossible, or would impose a preponderant burden of delay, trouble and expense, admit the writing. Refusing it could remove the only information available on a matter that decides the cause.¹¹

He did not overlook what a relay allows someone to evade. He distinguishes the person who perceived an event from the person who reports that it was described. The latter can tell the truth about what was said while the earlier account is false. But the remedy for danger need not be silence. Admission gives a judge something to consider; it does not oblige the judge to believe it. Bentham even notices the damage an exclusion can do to a conversation: remove what one participant said and the other’s words may become unintelligible. An effort to keep uncertain information from judgment can deprive judgment of the means to understand information it has accepted.¹²

I find the cost in this argument harder to dismiss than its confidence in the judge. Calling for better evidence sounds like an improvement until one asks what obtaining it requires, who can afford it and whether the interval leaves anything to decide. The unavailable witness may have made the only observation. An existing account may be the only way to recover a sequence of events. A demand that every link be reproduced in its strongest form can give the advantage to the party best able to wait. There is no reason to expect that party always to be the one with the truer account.

Bentham’s response nevertheless leaves the question pressed by Crawford: who may decide that the available assurances are an adequate substitute for a protected procedure? Confidence in a judge’s ability to discount weak evidence does not answer the objection to giving that judge the choice. Nor can a general preference for receiving information decide whether the State has met the conditions for using it to convict. The conflict survives a recognition of everyone’s good intentions. We need the work of absent people; we also need institutions whose appetite for a conclusion is constrained by more than their assessment of how likely it is to be correct.

Evidence law accommodates reliance in arrangements more particular than either a universal demand for presence or a universal license to repeat. Consider the federal business-records exception. Its foundation can be supplied by a custodian or another qualified witness, or through a permitted certification. The record’s original maker need not appear. The proponent must show that the record came from someone with knowledge, was made near the time and belonged to the activity’s regular way of keeping records. An opponent can defeat this route by showing untrustworthiness in the source or preparation. This route asks whether a way of making and keeping records supports their use; it does not imagine that the custodian personally performed every transaction.¹³

Other routes preserve other things. Former testimony can carry an earlier opportunity to develop a witness’s account, but the federal exception requires unavailability and the specified opportunity and similar motive—not simply an oath somewhere in the past. The exceptions for statements made under belief of imminent death and for statements against interest have their own conditions. Neither death nor exposure is a general certificate of truth. And fitting a hearsay exception does not by itself dispose of a confrontation objection. The legal distinctions are worth keeping because they prevent a useful accommodation from becoming an indiscriminate one.¹⁴

Possession of a sample, preservation of the document describing it, evidence that this is the document it purports to be, and grounds for accepting what the document says are separate achievements. We can authenticate a record and still dispute its account. We can preserve an account faithfully and discover that its author misinterpreted an observation. We can also establish a fact from new evidence after its first recorder is gone. The absence of an original witness does not dictate ignorance; it changes what work can still be done and which claims that work can sustain.

What the file can carry

Much of that work is within the reach of existing technical systems. A representation can preserve which document an observation came from, which operation transformed it and who acted for whom. PROV-O, the W3C’s provenance ontology, expressly distinguishes entities, activities, agents and derivations. Its account of delegation includes authority and responsibility for an activity, with some responsibility retained by the principal. It supplies more than a space for a source’s name. It can make the relation between an earlier artifact and a later result available for inspection.¹⁵

Such a record could identify which file an expert consulted, which results were selected and which analysis was performed, helping another reader separate a repeated conclusion from additional work. Preserving an earlier version can make a later correction intelligible. Checking a received file against an authenticated hash can supply strong evidence that its bytes match that reference; it cannot establish that the experiment described in those bytes occurred. The provenance account, in turn, states a history of production and use. An institution may have sound grounds for accepting that history, but the representation does not create those grounds merely by expressing it.¹⁶

There is no need to answer this with an infinite regress of demands for more metadata. An inquiry can stop on adequate grounds for its purpose. A laboratory’s procedures, independent records, a repeat test or corroborating evidence can provide such grounds in different circumstances. What matters is that the inquiry has somewhere to go when a consequential premise is disputed. A source reference may lead to the underlying observation; a retained calculation may allow the result to be checked; an accountable professional may identify an assumption that the summary concealed. These are different responses to different questions. They are not improved by describing all of them as verification and leaving the recipient to discover which occurred.

Suppose a recommendation draws on a patient’s structured allergy record, while a relevant report is already present in a specialist’s note that the process does not consult. Recovering that note can change which evidence governs the recommendation. The document’s availability and its use in the decision are distinct facts. The note may itself require interpretation or further checking; finding it does not turn every statement in it into a diagnosis. The recommendation can now be assessed against a piece of the patient’s recorded history it previously ignored.¹⁷

That operation differs from generating another explanation of the original recommendation. The explanation may identify the omission, and then it has helped. If it merely restates the result more persuasively, it has supplied no evidence about the omitted report. The same distinction applies to a human expert. A new speaker can bring new judgment without new observation, or new observation without resolving the old dispute. The recipient must be able to tell which contribution it is relying upon.

Here the criminal trial supplies a demanding instance of a broader institutional problem, not a rule to be exported unchanged. A research group, a lender and a court have different purposes, burdens and procedures. They may have good reasons to accept different forms of support. None gains an additional investigation simply by adding another competent voice to the same assertion. Repetition can make a claim easier to encounter and harder to distinguish from a finding; it need not make the claim better established.

The receiving institution must substantiate the premises material to the consequence it proposes, whether through adequate reliance on prior work or through grounds newly obtained. Where a challenge puts one of those premises in issue, a defense of the interpreter’s qualifications is not an answer to a question about an unexamined event. The institution must make the relevant support open to the kind of challenge the use requires, find a sufficient alternative or refrain from treating that premise as established. Exclusion is one possible result. A narrower conclusion or another investigation may be another. What it cannot do is transfer the unanswered question to a person whose name now stands above the conclusion and declare it answered.

Longoni could explain a chemical test. The notes could tell him that Rast had performed it. When asked whether it had been done, he could point back to what she wrote. That return to the record was the boundary of the knowledge he brought about the event. The prosecution wanted to carry the assertion further, into a verdict. It was for the prosecution to supply what that further use required.

Source notes

Footnotes

1. Smith v. Arizona, No. 22–899, 21 June 2024, official slip opinion, majority pp. 7–9, especially p. 9, citing Appendix to Petition 26a–27a. The Court’s account and reproduced trial extracts are the principal record used here. Rast’s departure is unexplained; it is not treated as misconduct or as a finding of legal unavailability. The State’s quoted expectation is its description of the proposed testimony, not evidence of Longoni’s private intentions. ↩

2. Smith, majority pp. 8–9, 15–16, reproducing App. 39a–42a, 46a. Eight items were tested; four were the subject of Longoni’s trial opinions. The account of Item 26 is confined to the methods and answers reproduced there. No specimen inspection, additional laboratory detail or continuous reconstructed laboratory scene is claimed. ↩

3. Smith, majority pp. 11–19. The holding concerns conveyed statements that support the expert opinion only if true. It does not prohibit every expert reliance on another person’s work. See especially p. 18 for the useful testimony the majority expressly preserves, and p. 19 n. 5 for its treatment of ordinary expert materials such as scientific literature. ↩

4. Smith, majority pp. 19–22; Thomas, concurring in part, pp. 1–3; Gorsuch, concurring in part, pp. 1–3. Thomas favors formality and solemnity rather than primary purpose; Gorsuch questions the need for this guidance and explores alternative readings without settling them. The majority remanded both testimonial status and the disputed forfeiture question. No subsequent proceedings are described. The dispute about notes versus report is preserved at majority pp. 20–21. ↩

5. Smith, Alito, joined by Roberts, concurring in the judgment, pp. 1–7. The eighty-three-page hypothetical and fourteen-page objection are reported at p. 5, citing Treadwell v. Nickel, 194 Cal. 243, 266 (1924). This is Alito’s documented historical comparison, not a new examination of the Treadwell trial transcript. His discussion of inherited professional knowledge at p. 2 quotes Wigmore’s 1904 treatment, §665(3). ↩

6. Smith, Alito concurrence pp. 7–12; Federal Rules of Evidence 703, 705 and 105. The federal rules are his comparison, not a substitution for Arizona’s governing rules. See his n. 1 disclaiming an interpretation of Arizona law. The official Federal Rules of Evidence, 2025 edition supplies the federal comparison used here. Rule 703 requires reasonable field-specific reliance and applies the substantially-outweighs test to the proponent’s disclosure of otherwise inadmissible basis material; Rule 705 allows compelled disclosure on cross-examination. ↩

7. Smith, Alito concurrence pp. 12–14. He agrees that Longoni asserted particular testing events as facts without personal knowledge and would vacate and remand on that ground. His objection to the majority’s broader reasoning is not an endorsement of all Longoni’s testimony. The majority’s preserved uses of forensic expertise are at p. 18. ↩

8. David Jardine, Criminal Trials, vol. I (London: Charles Knight, 1832), pp. 400, 434–435; digitized edition. The narrative marks Raleigh’s explanation as his argument. “Let Cobham be here” is the demand quoted in Crawford, 541 U.S. at 44, from Howell’s State Trials, vol. II, pp. 15–16. The exchange with Howard and Cecil is reported by Jardine at p. 435. This edition compiles reports; it is not a contemporary verbatim transcript made by Jardine. ↩

9. Jardine, vol. I, p. 400 editorial note and pp. 436–437 note. The editor identifies several reports and assesses their reliability; his inserted correspondence carries an express uncertainty about its use at Raleigh’s trial. The account does not infer that all printed documents were read to the jury. ↩

10. Crawford v. Washington, 541 U.S. 36 (2004), pp. 44, 50–51, 61–62, 68. The opinion’s constitutional reading of Raleigh is distinguished from the historical event and from the truth of the accusation. The general rule for testimonial statements of an absent witness requires unavailability and an earlier opportunity for cross-examination; Crawford also discusses forfeiture by wrongdoing and leaves a question concerning historical dying declarations. Neither the exceptions nor criminal confrontation doctrine are generalized into a universal institutional test. The confrontation provision was ratified with the Bill of Rights in 1791. ↩

11. Jeremy Bentham, Rationale of Judicial Evidence, in The Works of Jeremy Bentham, John Bowring ed., vol. VII (Edinburgh: William Tait, 1843), Book VI, ch. II §1, rules 1–2 and their reasons. The inspected version is the Online Library of Liberty v6.0 PDF, pp. 192–194 of that PDF. “Trouble” in the prose paraphrases Bentham’s “vexation”; his grounds include both physical impossibility and preponderant inconvenience. These are not page references to the 1827 Mill edition. ↩

12. Bentham, same edition, Book VI, ch. III, PDF pp. 208–210; ch. IV, pp. 211–213; ch. V opening, p. 214. His account distinguishes testimony that an assertion was made from the truth of the assertion, considers fabrication, and resists exclusion as the ordinary answer to a danger of deception. ↩

13. Federal Rules of Evidence, 2025 edition, Rule 803(6), clauses (A)–(E). Unavailability is not required under Rule 803. The permitted certification routes are Rule 902(11), 902(12), or a statute. Rule 803(6)(E) concerns an opponent’s showing that the source, method or circumstances indicate a lack of trustworthiness. The paragraph explains a federal route for regular records, not the admissibility of Rast’s particular records. ↩

14. Federal Rules of Evidence, 2025 edition, Rule 804(a), (b)(1)–(3). Former testimony requires an earlier trial, hearing or lawful deposition and an opportunity and similar motive for direct, cross- or redirect examination by the specified party (or, in a civil case, predecessor in interest). Rule 804(b)(2) concerns a prosecution for homicide or a civil case and a statement about the cause or circumstances of believed imminent death. Rule 804(b)(3) includes proprietary, pecuniary and liability interests; the specified criminal use of a statement against penal interest requires corroborating circumstances clearly indicating trustworthiness. The inspected edition includes the amendment effective 1 December 2024; it is not dated to the earlier Smith decision. Rule 901(a) separately states the authentication threshold. The chapter preserves these distinctions without treating any exception as conclusive truth or an automatic answer to a constitutional claim. ↩

15. Timothy Lebo, Satya Sahoo and Deborah McGuinness, eds., PROV-O: The PROV Ontology, W3C Recommendation, 30 April 2013, §§3–4, especially wasGeneratedBy, wasDerivedFrom, wasAttributedTo, actedOnBehalfOf and the qualified forms. Delegation includes authority and responsibility for an activity, with some responsibility retained by the principal; the vocabulary does not specify every legal allocation. The distinction between an expressed provenance relation and evidence supporting that relation is the chapter’s analysis. ↩

16. A collision-resistant hash comparison against an authenticated reference supports byte correspondence under its assumptions. It establishes neither creation-time truth nor a universal chain of institutional authority. The example of describing an expert’s file use is a proposed application of existing provenance capabilities, not a claim that the Smith laboratory used PROV-O. This proposed application supplies no implementation evidence, measured performance result or protocol guarantee. ↩

17. This is an explicitly hypothetical omission from the governing representation. It describes no patient, deployed system, drug recommendation or verified adverse event. Recovering the document can change the evidence considered; the document’s own content and the resulting recommendation may still need assessment. The hypothetical does not establish that tokens or tables necessarily discard provenance or that every omission has this repair. ↩