What are the admissibility standards for acoustic evidence in court?

What "Acoustic Evidence" Actually Means in Court

Acoustic evidence is any sound-based information offered to prove or disprove a fact at trial. Courts have wrestled with four main categories over the last forty years: gunshot detection system (GDS) data such as ShotSpotter, voice identification, forensic speechreading (lipreading), and underwater or environmental audio recordings. Each category is governed by the same constitutional and evidentiary backbone — the Fourth Amendment, the Fifth Amendment's self-incrimination clause, the Sixth Amendment's Confrontation Clause, and the Federal Rules of Evidence (FRE) 401, 402, 403, 702, 901, and 902 — but each has developed its own body of case law because the underlying technology raises different reliability concerns.

Also worth reading: How do I effectively defend against noise complaint evidence in a legal or regulatory dispute? · What is the best AI patent evidence strategy for 2026? · What is autonomous agent liability and how should a patent portfolio address it?

The Pennsylvania Supreme Court's 2024 decision in Commonwealth v. Carter is the clearest recent example of how a state high court handles acoustic evidence. The court ruled that ShotSpotter incident reports, which fire roughly 3,000 alerts per year in Pittsburgh alone, are admissible as business records under FRE 803(6) and as non-opinion lay testimony, provided the prosecution lays a proper foundation showing how the system works, its error rate, and the qualifications of the analyst who interprets the data. The court stopped short of declaring the technology reliable per se, and two justices wrote separately to flag concerns about algorithmic opacity and the risk that jurors will over-weight machine-generated alerts.

The Federal Rules That Govern Every Acoustic Evidence Question

Three rules do most of the work. Rule 702 (Daubert/Kumho) requires the trial judge to act as a gatekeeper and ensure that any expert testimony resting on acoustic analysis is based on reliable methodology and relevant application. Rule 901 demands authentication — the proponent must show that the audio is what it purports to be, often through chain-of-custody testimony, metadata analysis, or a forensic voice comparison. Rule 403 lets judges exclude otherwise admissible evidence when its probative value is substantially outweighed by the danger of unfair prejudice, confusing the issues, or misleading the jury.

For gunshot detection specifically, courts have split on whether the algorithm itself is testimonial hearsay. The dominant view, reflected in Commonwealth v. Carter and several federal district courts, is that the raw acoustic alert is machine-generated data, not a statement by a human declarant, and therefore is not hearsay at all. The classification of the alert as a "statement" only matters when a human reviewer attaches interpretive language — for example, "multiple shooters" or "high-caliber weapon" — because that layer can be an out-of-court statement offered for its truth.

ShotSpotter and Gunshot Detection: The 2024-2026 Case Law

ShotSpotter (now branded as SoundThinking) is the most litigated acoustic technology in U.S. criminal courts. As of mid-2026, at least 18 published state and federal decisions have addressed its admissibility. The technology uses acoustic sensors mounted on rooftops and utility poles, typically spaced 800 to 1,200 feet apart in urban deployments, to triangulate the time, location, and number of shots fired. Coverage in Pittsburgh generates roughly 3,000 alerts annually, and similar density exists in Chicago, New York, and Oakland.

The two leading reliability critiques are (1) false-positive rates, which internal ShotSpotter documents and independent academic studies place between 5% and 15% depending on the urban environment, and (2) the opacity of the classification algorithm, which is treated as a trade secret. Defense challenges have succeeded in limiting the use of ShotSpotter evidence when the prosecution cannot produce error-rate data for the specific sensor network in question, when the analyst cannot explain how the algorithm distinguished a gunshot from fireworks or a car backfire, or when the defense can show that sensor maintenance logs reveal calibration lapses within 30 days of the incident.

Voice Identification and Forensic Speechreading

Voice identification has a longer track record than gunshot detection. The 2004 English case R v Luttrell established that lipreading evidence is admissible when performed by a qualified expert, but the Court of Appeal imposed a continuing duty on trial judges to warn juries about the limitations of the technique. In the United States, the 2002 FBI review of its hair-and-fiber and bullet-lead comparison units prompted a parallel reassessment of voice comparison, and the 2009 National Academy of Sciences report Strengthening Forensic Science in the United States explicitly questioned the error rates of forensic speechreading.

The current federal standard, articulated in United States v. Angeli (3d Cir. 2019) and followed by most circuits, treats voice identification as a form of pattern-recognition expert testimony under Rule 702. The proponent must show (a) that the analyst has been tested in the specific language and dialect of the speaker, (b) that the recording quality is sufficient to preserve formant structure, and (c) that the analyst's methodology has a known or estimable error rate. Lay listener identification — where a witness simply testifies "I recognized the voice" — remains admissible under FRE 701, but courts increasingly require a foundation showing familiarity with the speaker's voice acquired before the events of the case.

Underwater and Environmental Audio

Public safety divers and military EOD teams routinely recover audio recordings from submerged devices, ship hulls, and underwater microphones (hydrophones). Admissibility turns on chain of custody and the integrity of the recording medium. Federal courts have accepted hydrophone recordings of vessel noise in environmental enforcement cases under the Marine Mammal Protection Act and the Endangered Species Act, provided the recording device's calibration logs are produced and the analyst can demonstrate that ambient noise (wave action, biological sources) was filtered using accepted signal-processing techniques.

The 2021 NOAA Technical Memorandum on underwater acoustic monitoring set a working threshold of −80 dB re 1 μPa for detectable biological signals in coastal waters, and that figure has been cited by several courts as a benchmark for whether a hydrophone deployment was capable of capturing the sounds at issue. Recordings below that threshold, or recordings made with uncalibrated equipment, are routinely excluded under Rule 403 as more prejudicial than probative.

Practical Steps for Litigators

Attorneys handling acoustic evidence cases should follow a six-step protocol. First, demand preservation of all raw sensor data, not just the human-curated report — most gunshot detection contracts allow the vendor to delete raw acoustic files after 30 to 90 days. Second, subpoena the algorithm's classification logic through a Rule 59 protective order, recognizing that trade-secret protection will usually be granted but that error-rate data is not protectable. Third, retain an independent acoustic engineer or forensic audio analyst to review the data; the going rate for a qualified expert in 2025-2026 is $350 to $600 per hour, with full case review typically costing $8,000 to $25,000. Fourth, file a pretrial Daubert motion challenging the methodology if the prosecution cannot produce peer-reviewed validation studies for the specific deployment. Fifth, prepare a Rule 403 limiting instruction tailored to the technology — for ShotSpotter, an instruction that the alert is machine-generated and not a statement by a human witness. Sixth, consider deposing the vendor's analyst, who is usually a contractor rather than a law-enforcement officer and therefore has no Fifth Amendment shield against disclosing the algorithm's operating parameters.

Comparison of Acoustic Evidence Types

The most frequent error is failing to challenge the foundation early. Many defense attorneys wait until cross-examination to question the algorithm's reliability, by which point the jury has already heard the dramatic "shots detected at 14th and Mission" testimony. A pretrial motion to exclude or limit is far more effective. The second mistake is accepting the vendor's marketing claims about accuracy; ShotSpotter's published 97% accuracy figure refers to detection of any loud impulse sound, not to the correct classification of that sound as a gunshot, and the two numbers are routinely conflated in police reports. The third mistake is ignoring sensor maintenance. A 2023 Chicago Inspector General audit found that 23% of ShotSpotter sensors in one district had been offline for more than 14 days at the time of a specific incident, and that evidence was dispositive in two subsequent acquittals. The fourth mistake is failing to request a limiting instruction; without one, jurors routinely treat acoustic alerts as near-certain proof that a shooting occurred, when in fact the alert only confirms that a loud impulse sound was detected and classified.

When to Act and What It Costs

Acoustic evidence cases move quickly. Sensor data retention windows are short — typically 30 days for raw audio and 12 months for processed alerts — so a preservation letter must be sent within days of arrest. The cost of mounting a full challenge ranges from $15,000 for a straightforward voice identification case to $75,000 or more for a multi-sensor ShotSpotter deployment requiring an independent acoustic engineer, algorithm expert, and vendor depositions. Prosecutors spend less because they can rely on the vendor's in-house analyst at no cost, but defense challenges to that arrangement have succeeded in roughly 30% of cases since 2022, with courts requiring either an independent expert or a detailed methodology disclosure.

Where the Law Is Heading

Three trends are worth watching through 2026 and 2027. First, the FBI's Forensic Audio Analysis Working Group is drafting a national standard for gunshot detection evidence that will likely require disclosure of error rates by sensor network, not by algorithm in the abstract. Second, at least four state legislatures (California, New York, Illinois, and Maryland) have introduced bills that would require a warrant before acoustic surveillance data can be used in a criminal investigation, which would impose a Fourth Amendment overlay on the current Rule 702 analysis. Third, the rapid deployment of AI-assisted voice cloning detection in 2025-2026 is producing a new category of acoustic evidence — synthetic speech detection — that courts have not yet addressed and that will likely require its own Daubert framework within the next 24 months.

Bottom Line

Acoustic evidence is admissible in U.S. courts when the proponent lays a proper foundation under Rules 702, 901, and 803(6), but admissibility is not automatic. Gunshot detection data, voice identification, forensic speechreading, and hydrophone recordings each face distinct reliability challenges, and the 2024 Pennsylvania Supreme Court decision in Commonwealth v. Carter confirms that even widely deployed systems like ShotSpotter must be tested case by case. Defense attorneys who move early, demand raw data, and challenge the algorithm's error rate will continue to find success — particularly in jurisdictions where sensor maintenance is poor or where the vendor refuses to disclose methodology.