Expert and forensic evidence

Grace Liu had examined latent fingerprints for fourteen years, and she had testified more than sixty times. She still prepared for the Bayview trial as if it were her first. She reread her bench notes, the photographs of the glass in place, the enlargements of the lift, and the record of her comparison. She checked that the verification by her colleague, Raj Varma, had been done blind, without access to her conclusion. She reviewed the questions she expected Marchetti to ask, about error rates, about contextual bias, and about what a fingerprint can and cannot say about when it was left. 'Most of my testimony,' she told Desjardins's class, 'is about the limits of what I know.'

Desjardins tells the students that expert evidence is the area of evidence law where police have most often contributed to wrongful convictions, not through dishonesty but through overconfidence. A forensic examiner who overstates a conclusion, or a court that accepts a technique without asking how reliable it is, can turn a weak case into a conviction. This lesson explains when expert evidence is allowed, how courts test it, and what officers and forensic specialists must do to make it trustworthy.

The opinion rule and lay opinion

Witnesses generally testify to facts they perceived, not to opinions or inferences drawn from those facts, because drawing inferences is the job of the trier of fact. But the line between fact and opinion is not sharp, and the law allows ordinary witnesses to give opinions on matters within common experience, such as a person's approximate age, speed, sobriety, emotional state or identity, where the opinion is a compact way of describing what they perceived. Ohio Evidence Rule 701 allows a lay witness to give opinions that are rationally based on the witness's perception and helpful to a clear understanding of the testimony or the determination of a fact in issue. Mrs. Shaw's statement that the man was 'tall' was a lay opinion of this kind. So was Okeke's testimony that Brennan 'appeared calm'.

When expert opinion is admissible

Expert evidence is opinion evidence on matters outside ordinary knowledge, given by someone with special training or experience. It is an exception to the opinion rule, and because it can be so persuasive, courts treat it carefully. In Daubert v. Merrell Dow Pharmaceuticals (1993), the United States Supreme Court held that trial judges must act as gatekeepers, ensuring that expert testimony is both relevant and reliable, considering factors such as whether the method can be and has been tested, whether it has been subject to peer review, its known or potential error rate, the existence of standards controlling its operation, and its general acceptance in the relevant scientific community. Kumho Tire v. Carmichael (1999) applied the gatekeeping duty to all expert testimony, including technical and experience-based fields. The Supreme Court of Ohio adopted a similar approach in Miller v. Bike Athletic Co. (1998). Ohio Evidence Rule 702, amended effective July 1, 2024 to mirror the 2023 federal amendment, requires the proponent to show the court that it is more likely than not that the witness is qualified, that the testimony is based on reliable scientific, technical or other specialized information, and that the opinion reflects a reliable application of the principles and methods to the facts of the case. Under Criminal Rule 16(K), an expert must prepare a written report summarizing the testimony, findings, analysis, conclusions and qualifications, which must be disclosed at least 21 days before trial, or the testimony may be excluded.

Qualifying the expert

At trial, the party calling an expert first asks the court to qualify the witness in a specific field. The witness describes their training, experience, certifications, proficiency testing and previous testimony, and the other side may cross-examine on qualifications before the judge rules. Kwan asked that Liu be qualified as an expert in latent print examination. Marchetti did not contest her qualifications at the hearing, but asked about her proficiency testing and the laboratory's accreditation and quality procedures, to lay the groundwork for cross-examination on weight. Desjardins notes that the field of expertise must be defined precisely, because an expert may give opinions only within it. A fingerprint examiner can say whose finger left a print; she cannot say when or why it was left, or whether the person was the burglar.

Fingerprint evidence and its limits

Fingerprint examiners use a method called ACE-V: analysis, comparison, evaluation and verification. In analysis, the examiner studies the latent print to decide whether it has enough clear detail to be compared. In comparison, she compares it side by side with known prints. In evaluation, she reaches a conclusion: identification, exclusion or inconclusive. In verification, a second examiner repeats the process. The method depends on the examiner's judgment; there is no fixed number of matching features that defines an identification, and most agencies abandoned numerical standards years ago.

Major reviews have criticized the way fingerprint evidence was sometimes presented in the past. The 2009 report of the United States National Academy of Sciences on forensic science, and the 2016 report of the President's Council of Advisors on Science and Technology, concluded that latent print examination is a valid method, but that its error rate is not zero, that examiners can be influenced by contextual information about the case, and that claims of absolute certainty or zero error are not scientifically supportable. Studies of examiners working under controlled conditions have found false identifications to be rare but real. The best-known modern error was the 2004 Madrid train bombing case, where FBI examiners wrongly identified a latent print as belonging to Brandon Mayfield, an Oregon lawyer, and verifiers confirmed the error because they knew the first conclusion. Many American laboratories, including those in Ohio, are now accredited to international standards and have adopted blind verification and standard reporting language. The reforms include documenting each step, verifying blind, limiting the examiner's exposure to unnecessary case information, and reporting conclusions in measured language.

# An automated search returns a ranked candidate list; the score is not an identification.
candidates = [
    ("candidate A", 3120), ("candidate B", 2985), ("candidate C", 2410), ("candidate D", 2366),
    ("candidate E", 2201), ("candidate F", 2150), ("candidate G", 2098), ("candidate H", 1990),
]
comparisons = {"candidate A": "identification", "candidate B": "exclusion"}
print("automated search: ranked candidates by similarity score")
for name, score in candidates:
    result = comparisons.get(name, "not compared")
    print(f"  {name}  score {score}  -> examiner: {result}")
ids = [n for n, r in comparisons.items() if r == "identification"]
print("identifications by the examiner:", len(ids), "| verified blind by a second examiner:", True)

automated search: ranked candidates by similarity score
  candidate A  score 3120  -> examiner: identification
  candidate B  score 2985  -> examiner: exclusion
  candidate C  score 2410  -> examiner: not compared
  candidate D  score 2366  -> examiner: not compared
  candidate E  score 2201  -> examiner: not compared
  candidate F  score 2150  -> examiner: not compared
  candidate G  score 2098  -> examiner: not compared
  candidate H  score 1990  -> examiner: not compared
identifications by the examiner: 1 | verified blind by a second examiner: True

The automated search did not identify anyone. It ranked candidates from the database by how similar their recorded prints were to the latent. Liu compared the latent with the known prints of the first two candidates. She excluded the second and identified the first, Brennan, then documented her reasons. Varma, given only the latent and a set of known prints and told nothing about her result, reached the same conclusion. Desjardins asks why Liu compared more than one candidate. Jesse Morningstar says it shows she was testing the result rather than confirming it, and that a high score does not mean a match. That is exactly the point.

Testimony about the Bayview print

Liu testified that the latent print on exhibit 4A, lifted from the inside face of a glass shard that remained in the lower window frame, was made by the right middle finger of Shane Brennan. She explained the method, showed enlargements of corresponding features, and said that her conclusion was verified by a second examiner. On cross-examination, Marchetti asked whether fingerprint identification was infallible. Liu said no. She asked whether Liu could say when the print was left. Liu said she could not. She asked whether the till drawer print had been identified. Liu said it was not suitable for comparison, and that she had reported that. Marchetti asked whether Liu had known, when she did the comparison, that Brennan had been found with stolen property. Liu said she had not; she had received the lift and the candidate list from the system and nothing else.

The question of when the print was left mattered. A fingerprint on the outside of a shop window could have been left by anyone at any time. But this print was on the inside face of the glass, on a rear window that opened into the pharmacy's back storage room, where customers never went. Whitaker testified that the window had been installed in 2023, was cleaned every month, and could not be reached from inside without moving shelving. Liu could not say when the print was made, and properly refused to. But Whitaker's evidence gave the court a basis to infer that the only reasonable explanation for Brennan's fingerprint on the inside of that glass was that he had reached through the broken window. The expert evidence and the ordinary evidence together proved what neither proved alone.

Forensic reports and the Confrontation Clause

The United States Supreme Court has held that a forensic report prepared for use at trial is testimonial, so the analyst who made it must ordinarily testify: in Melendez-Diaz v. Massachusetts (2009) for drug certificates, in Bullcoming v. New Mexico (2011) for a blood-alcohol report, and in Smith v. Arizona (2024), where it held that a substitute analyst cannot simply relay the absent analyst's findings as the basis for their own opinion. The examiner who did the work should expect to testify. Liu and Varma both attended, though only Liu was asked substantive questions.

Other forensic and technical experts

The same principles apply to every kind of expert police rely on: DNA analysts, firearms examiners, toxicologists, collision reconstructionists, digital forensic examiners and medical examiners. Some fields rest on strong science, such as single-source DNA analysis; others, such as bite-mark comparison, have been widely discredited. Reviews by the National Registry of Exonerations and others have found that false or misleading forensic evidence contributed to a significant share of known wrongful convictions in the United States. Officers who commission expert work should give the expert the materials needed and nothing that might bias the result, should never pressure an expert toward a conclusion, and should disclose every report, including those that are unhelpful or inconclusive.