Abstract
A long witness list suggests to jurors and observers that a fact has been confirmed many times over. That suggestion is often mistaken. This paper distinguishes two structures in which multiple witnesses appear at trial. In corroboration, several witnesses attest to the same matter, each capable of confirming or contradicting the others, and their agreement increases assurance. In sequence, single witnesses attest to successive links in a chain, each speaking only to his own link, and the addition of links decreases the overall assurance of the chain even as it lengthens the witness list. The paper sets out the conditions that genuine corroboration requires, shows by simple arithmetic why sequence weakens with length, identifies common forms of apparent corroboration that fail those conditions, and describes practices that introduce true corroboration within a chain. It applies the distinction to five composite case examples and concludes with implications for courts, laboratories, attorneys, and jurors.
1. Introduction
White Paper 1 classified evidence by its foundation load, the number of persons whose testimony or certification is required before a result can be presented. It observed that the heaviest loads arise where physical evidence, testimonial reports, and divided laboratory workflows coincide, as in DNA typing, and that the resulting procession of witnesses consists largely of persons who each speak to a single stage.
This paper takes up the question that such a procession invites. Does the appearance of many witnesses mean that the fact they support has been established with correspondingly greater certainty? The intuitive answer is yes. Ordinary experience teaches that a report confirmed by several people deserves more confidence than one resting on a single account. The law has long reflected that experience, requiring more than one witness for certain grave findings.[1]
The intuition holds only when the witnesses speak to the same thing. When they speak to different things arranged in a series, the arithmetic reverses. This paper names the two structures, states the conditions that separate them, and examines how the difference is obscured in practice.
2. Two Structures of Multiple Testimony
2.1 Corroboration
Corroboration is attestation by more than one witness to the same matter. For the purposes of this series, genuine corroboration requires three conditions.
The first is identity of subject. The witnesses must attest to the same event, act, or fact, not to adjacent or related ones. Two witnesses who each saw a different stage of a transaction have not corroborated one another as to either stage.
The second is independence of source. Each witness’s account must rest on his own perception or his own analysis, not on a common source shared with the other witnesses. Two witnesses who both read the same log entry have not independently observed the event it records.
The third is capacity for contradiction. Each witness must be in a position to have disagreed with the others had the facts been different. A second examiner who knows the first examiner’s conclusion before beginning, and whose role is understood as confirmation, is less able to contradict than one who examines the evidence without that knowledge.
When all three conditions are met, agreement among witnesses is informative in a way a single witness’s account cannot be. Each witness serves as a check on the others, and a false or mistaken account must survive comparison with an independent one.
2.2 Sequence
Sequence is attestation by single witnesses to successive links in a chain. The collecting officer testifies to collection, the property clerk to storage, the courier to transport, the laboratory intake technician to receipt, and each analyst to his own stage of processing. Each is ordinarily the only witness to his link.
The witnesses in a sequence do not check one another. The courier cannot confirm what the property clerk did in the property room, and the analyst performing amplification cannot confirm the quality of the extraction that preceded it except insofar as its output appears regular. The chain as a whole is established only if every link is established, and each link rests on one person’s account.
The analytical vocabulary of evidence scholarship has a term for inferences of this form. Chains of reasoning in which each step depends on the one before have been called cascaded or catenate inferences, and their treatment in the probabilistic literature on evidence makes clear that the strength of the conclusion at the end of such a chain cannot exceed, and ordinarily falls below, the strength of its weakest step.[2]
3. The Arithmetic of the Two Structures
A simplified model illustrates why the two structures behave differently. The figures below are illustrative, not empirical.
3.1 Sequence Weakens With Length
Suppose each link in a chain of custody and analysis is established with a reliability of 0.99, meaning that there is a one percent chance that the link failed in some way bearing on the result, through mislabeling, substitution, contamination, or an undetected procedural error. If the links fail independently, the probability that every link held is the product of their reliabilities.
| Number of links | Probability all links held |
|---|---|
| 1 | 0.990 |
| 5 | 0.951 |
| 10 | 0.904 |
| 15 | 0.860 |
| 20 | 0.818 |
Each added link lengthens the witness list by one person and lowers the probability that the chain is sound. A chain of twenty highly reliable links is considerably less certain than any one of them. The procession that appears to fortify the result is, in this respect, the measure of how many opportunities for failure it contains.
Two qualifications are necessary. First, not every failure affects the result; a mislabeled box that is corrected before analysis does no harm. Second, the failure probabilities of real links are neither known nor uniform. The point of the model is structural: in a sequence, adding links cannot raise the certainty of the whole and ordinarily lowers it.
3.2 Corroboration Strengthens With Number
Suppose instead that two witnesses independently attest to the same matter, and each has a ten percent chance of being mistaken. If their errors are independent, the chance that both are mistaken is one percent, and the chance that both are mistaken in the same direction, so that their accounts agree on a falsehood, is lower still. A third independent witness reduces the figure further.
The arithmetic depends entirely on independence. If the two witnesses share a common source of error, such as the same faulty record or the same suggestive prompt, their errors are correlated, and their agreement adds little beyond what one of them provides. This is why the independence condition in Section 2.1 is not a formality. It is the condition on which the entire value of corroboration rests.[3]
3.3 The Asymmetry
The two structures thus move in opposite directions as witnesses are added. Corroboration grows more certain as independent witnesses to one matter are added. Sequence grows less certain as single witnesses to additional matters are added. A trial observer who counts witnesses without asking which structure they form will systematically overestimate the assurance that sequence provides.
4. The Legal Tradition’s Recognition of the Difference
The distinction is not new. Where the law has required multiple witnesses as a condition of a finding, it has ordinarily specified that they attest to the same matter.
The Constitution provides that no person shall be convicted of treason except on the testimony of two witnesses to the same overt act, or on confession in open court.[4] The Supreme Court has emphasized that the requirement attaches to the overt act itself and that each element of the act must be supported by the testimony of two witnesses.[5] Two witnesses to different overt acts do not satisfy it.
The traditional rule for perjury similarly requires that falsity be proved by more than the uncorroborated oath of one witness, a rule the Court upheld in Weiler v. United States on the reasoning that one oath should not suffice to convict on the basis of another.[6]
The Hebrew Scriptures, from which the Anglo-American tradition drew much of its thinking about witness requirements, state the principle in terms of a single matter: at the mouth of two or three witnesses shall the matter be established.[7] The companion biblical volume in this suite treats that standard at length.
By contrast, the law of chain of custody treats the chain as a sequence and does not require corroboration of its links. Courts generally hold that the chain need not be perfect, that gaps ordinarily go to the weight rather than the admissibility of the evidence, and that the prosecution need not call every person who handled an item.[8] Some courts apply a presumption that official custodians handled evidence properly in the absence of evidence of tampering.[9] These rules are defensible as a matter of practicality, since requiring two witnesses to every custody transfer would be burdensome. But they confirm that a chain of custody is, in legal structure, a series of single attestations. The procession in a custody case is not the two-witness requirement in action; it is something else that resembles it.
5. Apparent Corroboration
Several common features of trial evidence present the appearance of corroboration while failing one or more of its conditions.
5.1 The Common Source
When several witnesses testify to a fact they learned from the same record, their testimony has a single source. A detective, a supervisor, and a prosecutor’s investigator who each describe the contents of the same property log are not three witnesses to the storage of the item; they are three reporters of one entry. The independence condition fails.
5.2 Non-Blind Verification
Many forensic disciplines require that a second examiner verify the first examiner’s conclusion. Verification is valuable, but its value as corroboration depends on whether the verifier works independently. Where the verifier knows the original conclusion, the research literature on forensic confirmation bias indicates that the verifier’s judgment may be influenced by it, reducing the capacity for contradiction.[10] A report reading “identification, verified” may therefore reflect one judgment and one review of that judgment rather than two independent judgments.
5.3 The Restated Result
A laboratory result may be restated by several witnesses: the analyst who produced it, a technical reviewer who checked the paperwork, a supervisor who signed the report, and an officer who received it. Each may testify accurately, and the jury may hear the result four times. But only the analyst attested to the analysis. The others attested to the regularity of documents or to receipt of the report. Repetition is not corroboration.
5.4 The Cumulative Officer
Several officers may be present at a scene, and each may testify to aspects of the search or seizure. Where they observed the same act, such as the discovery of an item in a particular location, their testimony may be genuine corroboration. Where each observed a different act, or where later officers arrived after the discovery and learned of it from the first, the testimony is sequence or common source presented in the form of multiple witnesses.
5.5 The Divided Pipeline
As described in White Paper 1, a pipeline workflow multiplies the analysts who may testify. Each speaks to a different stage. The jury may see four laboratory witnesses in a DNA case and reasonably infer that the result was checked four times. In structural terms, it was performed once, in four parts.
6. Corroboration Within Sequence
Sequence cannot be eliminated from forensic proof, since evidence must move from scene to laboratory to courtroom. But practices exist that introduce genuine corroboration at particular links, strengthening the chain where it would otherwise rest on a single attestation.
6.1 Dual Documentation of Transfers
A custody transfer documented independently by both the releasing and the receiving party, each generating its own record, allows the two records to be compared. If the item described, the time, and the condition agree, the transfer is attested from two sides. Electronic custody systems that timestamp both release and receipt serve the same purpose.
6.2 Tamper-Evident Packaging
A sealed package, bearing the initials of the person who sealed it and inspected upon opening by a person who confirms the seal was intact, allows the condition of the seal to corroborate the account of every custodian between sealing and opening. The intact seal is a physical witness to the intervening links. Its value is considerable precisely because it is independent of those custodians’ own testimony.
6.3 Independent Retesting and Split Samples
Where a sample is large enough to be divided, a portion retained for independent testing allows a second laboratory or examiner to repeat the analysis. Agreement between the two results is genuine corroboration of the analytical conclusion, satisfying identity of subject, independence of source, and capacity for contradiction. Where evidence is consumed in the original analysis, as frequently occurs with trace DNA, this form of corroboration is foreclosed, and the analytical links revert to sequence.
6.4 Blind Verification
Verification performed without knowledge of the original conclusion restores the capacity for contradiction that non-blind verification weakens. It converts a review into a second independent judgment.[11]
6.5 Convergence of Independent Lines of Evidence
At the level of the ultimate fact, corroboration may be supplied not by additional witnesses within one chain but by separate chains that arrive at the same conclusion. A DNA result placing the defendant’s biological material at a scene, a cell site analysis placing his phone in the area, and a surveillance recording showing a person of his description nearby are three chains with different sources of error. If they agree, their agreement is informative in a way the length of any one chain is not. Evidence scholarship has treated such convergence as a distinct and important form of corroboration.[12] Its value, like all corroboration, depends on the independence of the lines. Where a detective used one line to direct the gathering of another, the lines may share a source of error.
7. Case Examples
The following examples are composites constructed for illustration. They are not drawn from particular trials, and their details are simplified to isolate the distinction under discussion.
Example A: The Controlled Substance Case
The prosecution calls seven witnesses concerning a bag of powder: the arresting officer who seized it, a second officer who transported it, the property clerk who logged it, a courier who delivered it to the laboratory, the intake technician, the analyst, and the analyst’s supervisor. The jury hears seven witnesses and a positive result for a controlled substance.
In structural terms, the first six witnesses attest to six successive links, each on his own account. The supervisor attests to the review of paperwork, not to the analysis. No link is corroborated. If the bag was sealed at seizure and the analyst confirms the seal was intact, the seal corroborates the intermediate links; if it was not sealed, or the analyst cannot recall, it does not. The seven witnesses establish a sequence of six links and one restatement.
Example B: The Split DNA Sample
A bloodstain large enough to divide is tested by the state laboratory, and a portion is released to a defense-retained laboratory, which tests it independently. Both laboratories report the same profile.
The analytical conclusion is genuinely corroborated. The two laboratories worked from the same stain but through independent processes with independent sources of error. The custody links before the division remain a sequence, and an error there, such as contamination at collection, would affect both portions equally. Corroboration thus strengthens the analysis without reaching back to the collection. The example shows that corroboration is specific to the link at which it occurs.
Example C: The Verified Latent Print
An examiner identifies a latent print as the defendant’s. A second examiner, given the case file including the first examiner’s conclusion, verifies the identification. The report states “identification, verified,” and both examiners testify.
The jury sees two witnesses agreeing. The capacity for contradiction of the second is weakened by prior knowledge of the first’s conclusion. The example presents apparent corroboration of uncertain value. Had the verification been blind, the agreement would be considerably more informative.
Example D: The Converging Location Evidence
A cell site analyst testifies that the defendant’s phone connected to a tower whose coverage included the scene during the relevant hour. Separately, a store’s surveillance recording, authenticated by a store employee, shows a vehicle registered to the defendant in a parking lot near the scene during the same hour. The two lines of evidence were gathered independently, the recording having been collected before the phone records were obtained.
The two lines converge on the defendant’s presence in the area. Their sources of error differ: the cell site analysis depends on network interpretation, and the recording depends on vehicle identification. Their agreement is genuine corroboration at the level of the ultimate fact, though neither line individually requires a lengthy procession.
Example E: The Three Officers
Three officers testify to the discovery of a weapon in a vehicle. On cross-examination, it emerges that one officer found the weapon, a second officer was standing beside the first and saw it removed, and the third arrived after the discovery and was told where it had been found.
The first two officers corroborate one another as to the discovery. The third officer’s testimony about location rests on a common source, the first officer’s statement. The jury heard three witnesses; the discovery was attested by two, and the location of the weapon, if the second officer saw only its removal, perhaps by one.
8. Implications
8.1 For Jurors and Observers
The number of witnesses is not a measure of assurance. The question to ask of any group of witnesses is whether they saw the same thing independently or whether each saw a different part. A procession establishes that a chain was documented. It does not establish that any link was confirmed.
8.2 For Attorneys
The distinction offers a disciplined approach to challenging or defending custody proof. Counsel evaluating a witness list may identify which links rest on a single attestation and whether any practice, such as a sealed package or dual documentation, supplies independent confirmation. Arguments about “gaps in the chain” are more persuasive when they identify a link resting on one person’s account with no physical or documentary corroboration.
8.3 For Laboratories and Agencies
Laboratories and law enforcement agencies can strengthen chains without lengthening them by adopting practices that introduce corroboration at vulnerable links: tamper-evident sealing, dual documentation of transfers, blind verification, and the preservation of retestable portions where samples permit. These practices are addressed as recommendations in White Paper 6.
8.4 For Courts
Jury instructions ordinarily tell jurors that the number of witnesses is not decisive and that they should weigh testimony rather than count it. Instructions that explain the difference between witnesses who confirm one another and witnesses who each describe a separate step would apply that general principle to the setting in which it is most often misunderstood. White Paper 6 proposes model language.
9. Limitations
The arithmetic in Section 3 assumes independent failures and fixed link reliabilities; real chains have correlated risks, such as a single careless laboratory, and reliabilities that cannot be measured precisely. The conditions of corroboration in Section 2.1 are matters of degree, and courts and juries must judge how independent two witnesses actually were. The composite examples simplify the documentation and circumstances of actual cases. The framework identifies what questions to ask of multiple testimony; it does not supply numerical answers for any given trial.
10. Conclusion
Multiple witnesses may stand in two very different relations to the fact they support. When they attest independently to the same matter, each capable of contradicting the others, their agreement multiplies assurance. When they attest singly to successive links, the chain they form is no stronger than its weakest link, and each added link adds an opportunity for failure. The courtroom presents both structures in the same outward form, a person taking the stand, and the difference between them is rarely made explicit.
The legal tradition has long required corroboration for its gravest findings and has specified that it must go to the same matter. Chain of custody proof, by contrast, is built on sequence, and the procession it produces resembles corroboration without being it. Recognizing the difference does not diminish the value of careful custody proof; it clarifies what that proof establishes and where it may be strengthened. The following paper turns to the constitutional line that determines which links in the chain must appear in person, and asks whether that line follows the reliability concerns this paper has described.
Notes
[1] See the constitutional two-witness requirement for treason and the traditional corroboration rule for perjury, discussed in Section 4.
[2] Schum (1994) develops the analysis of cascaded or catenate inference and its implications for the force of evidence at the end of a chain of reasoning. Anderson, Schum, and Twining (2005) present the same concepts in the context of legal evidence analysis.
[3] Cohen (1977) and Schum (1994) both treat independence as the condition on which the evidential force of corroborating testimony depends. Schum distinguishes corroboration, in which several sources report the same event, from convergence, in which reports of different events point to the same conclusion.
[4] U.S. Const. art. III, § 3, cl. 1.
[5] Cramer v. United States (1945).
[6] Weiler v. United States (1945).
[7] Deuteronomy 19:15; see also Deuteronomy 17:6 and Numbers 35:30.
[8] Melendez-Diaz v. Massachusetts (2009), p. 311 n.1; Giannelli (1983); Mosteller (2020).
[9] Giannelli (1983) discusses the presumption of regularity as applied to official custodians and the criticism it has drawn.
[10] Kassin, Dror, and Kukucka (2013) review research on forensic confirmation bias and propose procedural safeguards, including blind verification. See also Dror and Hampikian (2011) on contextual influence in DNA mixture interpretation.
[11] Kassin, Dror, and Kukucka (2013); National Research Council (2009) recommended research and procedural measures to reduce bias in forensic examination.
[12] Schum (1994); Anderson, Schum, and Twining (2005). Kadane and Schum (1996) apply these concepts in a detailed probabilistic analysis of the evidence in a historical criminal case.
References
Anderson, T., Schum, D., & Twining, W. (2005). Analysis of evidence (2nd ed.). Cambridge University Press.
Cohen, L. J. (1977). The probable and the provable. Clarendon Press.
Cramer v. United States, 325 U.S. 1 (1945).
Dror, I. E., & Hampikian, G. (2011). Subjectivity and bias in forensic DNA mixture interpretation. Science & Justice, 51(4), 204–208.
Giannelli, P. C. (1983). Chain of custody and the handling of real evidence. American Criminal Law Review, 20(4), 527–568.
Kadane, J. B., & Schum, D. A. (1996). A probabilistic analysis of the Sacco and Vanzetti evidence. Wiley.
Kassin, S. M., Dror, I. E., & Kukucka, J. (2013). The forensic confirmation bias: Problems, perspectives, and proposed solutions. Journal of Applied Research in Memory and Cognition, 2(1), 42–52.
Melendez-Diaz v. Massachusetts, 557 U.S. 305 (2009).
Mosteller, R. P. (Ed.). (2020). McCormick on evidence (8th ed.). West Academic.
National Research Council. (2009). Strengthening forensic science in the United States: A path forward. National Academies Press.
Schum, D. A. (1994). The evidential foundations of probabilistic reasoning. Wiley.
U.S. Const. art. III, § 3.
Weiler v. United States, 323 U.S. 606 (1945).
