Voice Notes and Recordings
Audio feels like the most direct evidence there is, which is exactly why it repays scepticism. Before a recording establishes anything about what was said, somebody has to establish where it came from, who is on it, and whether anything is missing from it.
The rule in short
A recording is authenticated by evidence of how it was made and by whom, identification of the voices on it, and confirmation that it is complete and unaltered. Transcripts are aids rather than evidence. Editing is easy and increasingly hard to detect, so provenance carries more weight than any examination of the audio itself.
Recordings arrive in files with an air of finality about them, and they are treated as though the only question is what they contain. Three earlier questions have to be answered first, and each of them is answered by evidence outside the audio rather than by listening to it.
The three questions before content
Where did it come from. Who made the recording, on what device, in what circumstances and why. An account of that is what turns an audio file into evidence, and a recording nobody can account for is a recording nobody has to accept.
Who is speaking. Identification of every voice on the recording, which is given by somebody familiar with the speaker and does not require any expertise. Familiarity acquired at any time is generally sufficient for this purpose.
Is it complete. Whether the recording covers the whole conversation or begins and ends at points somebody chose. A recording that starts mid-sentence invites the question of what preceded it, and the answer matters.
Only then, what it says. The content is assessed once those three are settled, and arguing about content before they are settled is how parties end up disputing a file nobody has established anything about.
Each is separate evidence. Provenance comes from the maker, identification from somebody who knows the voice, completeness from the maker and from the circumstances. None of the three is answered by the audio itself.
Authenticating a recording
A witness who made it. The most straightforward route: the person who pressed record explains when, where, why and what they did with the file afterward. That covers provenance and completeness together.
Or a witness who was there. Somebody present who can say that the recording fairly represents what they heard is an alternative where the maker is unavailable, and their evidence is about accuracy rather than about the file.
Or evidence about the system. Where a recording was made automatically by a business, evidence describing how the system operates serves the same function, on the reasoning set out in business records and the routine entry.
Distinctive contents can help. A recording containing information only particular people would know can be authenticated partly by its own contents, which is a recognized route where nothing else is available.
The original file matters. It carries creation times, device details and format characteristics that a re-encoded copy sent through a messaging application has discarded entirely.
| Question | Answered by | Not answered by |
|---|---|---|
| Where the recording came from | The person who made it | Listening to it |
| Who is speaking | Somebody familiar with the voice | The file itself |
| Whether it is complete | The maker and the circumstances | The absence of audible joins |
| What was said | The recording | The transcript |
Identifying voices
Familiarity is the ordinary test. Anybody who knows a person's voice can identify it, and the familiarity may have been acquired before or after the recording was made. No qualification is required for this.
The weight varies with the familiarity. A spouse identifying a voice heard daily for twenty years is in a different position from somebody who spoke to the person twice, and the difference goes to weight rather than admissibility.
Self-identification helps. Where a speaker names themselves on the recording, or refers to circumstances only they would know, the identification question becomes considerably easier to answer.
Expert comparison exists and is contested. Forensic voice comparison is available and is treated with more caution than most techniques, which is a matter for who qualifies as an expert.
Poor quality undermines everything. A recording in which speech is barely audible produces identification evidence that is little more than a guess, and honest witnesses should say so rather than straining.
A recording forwarded through a messaging application has been re-encoded, losing the creation data and the format characteristics that would show how it was produced. The file as it sits on the recording device is a different and much more informative object.
Transcripts, and what they are for
They are aids, not evidence. The recording is the evidence and the transcript helps people follow it. Where the two differ, the recording governs, and that should be stated whenever a transcript is produced.
Who prepared it matters. A transcript prepared by a party is a party's reading of the audio, and an independent transcriber's version is worth considerably more where any passage is contested.
Inaudible passages must be marked. Rather than guessed at, because a plausible guess presented as a transcription is the single thing most likely to discredit the whole document.
Speaker attribution is interpretation. Assigning lines to named speakers embeds an identification decision inside what looks like a neutral document, and that decision should be visible rather than assumed.
Timestamps make it usable. A transcript keyed to positions in the recording allows anybody to check a passage in seconds, which is what makes the aid genuinely useful.
Editing and completeness
Editing is easy. Removing a passage from an audio file requires no skill and free software, and a competent edit at a natural pause is extremely difficult to detect from the audio alone.
Which puts the weight on provenance. The account of how a recording was made and handled does more to establish its integrity than any examination of the waveform, and it is the part that can actually be tested.
Look at the file, not the sound. Encoding characteristics, file structure and creation data frequently reveal that a file was produced by re-encoding rather than by recording, as what metadata records sets out.
Selective recording is not editing. A person who started recording partway through a conversation has produced a complete file of an incomplete conversation, which is a different objection with a different answer.
Ask for everything from the device. The full set of recordings from the same period, in original form, which reveals both what was omitted and whether the produced file matches its neighbors.
Recordings persuade because they appear to remove the interpretive layer, and that is precisely the reason to examine them carefully. What a decision-maker hears is a file somebody produced, selected and handed over, and each of those steps involved a choice.
Provenance does most of the work. A recording whose making can be described in detail, by somebody who can be questioned about it, is far stronger than one that arrived from an unexplained source however clear the audio happens to be.
Identification is easier than people expect and needs to be pinned down anyway. Familiarity is the whole test, no expertise is required, and the strength of the evidence depends on how well the witness actually knows the voice.
Transcripts cause more trouble than they solve when their status is left unstated. They are aids to the recording, the recording governs where the two differ, and inaudible passages must be marked as inaudible rather than filled in.
On completeness, the honest question is what was not recorded rather than what was cut out. Selective recording is far more common than editing, and it is answered by asking for everything the device holds from the same period. Whether a recording could lawfully be made at all comes before any of this, and is dealt with in recording a Conversation.
Points to carry away
- Provenance, identification and completeness are three separate questions.
- A transcript is an aid to the recording rather than evidence itself.
- Voice identification can be given by anybody familiar with the voice.
- Editing is easy, so the account of how a recording was made matters most.
- Original files carry information that shared copies discard.
Questions readers ask
Who can identify a voice on a recording?
Anybody familiar with the speaker's voice, and the familiarity can have been acquired at any time, including after the recording was made. No qualification or expertise is required, and this is one of the more permissive rules in evidence practice. What varies is weight: somebody who has spoken with the person daily for years gives identification evidence of a different order from somebody who met them twice, and a decision-maker will weigh the two accordingly.
Is a transcript evidence of what was said?
No. The recording is the evidence and the transcript is an aid to following it, which should be stated expressly whenever one is produced. Where the transcript and the recording differ, the recording governs. That matters because transcripts embed decisions: which speaker said which line, what an unclear passage contains, and where one utterance ends and another begins. A transcript prepared by a party is that party's reading of the audio, not a neutral record of it.
How can an edited recording be detected?
Frequently it cannot, from the audio alone, because a competent edit made at a natural pause leaves nothing audible. What produces results is examining the file rather than the sound: encoding characteristics, structural inconsistencies and creation data that indicate re-encoding rather than direct recording. Beyond that, the productive approach is provenance and comparison — asking for the original file and for everything else recorded on the same device in the same period.
Sources
- Federal Rules of Evidence — Rule 901, Authenticating or Identifying Evidencelaw.cornell.edu
- Federal Rules of Evidence — Rule 1002, Requirement of the Originallaw.cornell.edu
- Federal Rules of Evidence — Rule 1003, Admissibility of Duplicateslaw.cornell.edu
- Federal Rules of Evidence — Rule 702, Testimony by Expert Witnesseslaw.cornell.edu
- Federal Rules of Evidence — Rule 803, Exceptions to the Rule Against Hearsaylaw.cornell.edu
- 18 U.S. Code § 2511 — Interception of Communicationslaw.cornell.edu
True Justice Record is a publication, not a law firm. This article states general rules and cites its sources; it is not advice about any particular case, and the law differs by state and changes over time.
More in Evidence That Lives on a Phone
When a Device Is Lost or Wiped
Where a device is gone, provider-held account data is unaffected, backups may capture an earlier state, and the other participants in any conversation hold their own copies. The circumstances of the loss then matter separately: an ordinary loss is neutral, while a wipe performed after a duty to preserve arose is treated as spoliation.
Location History Offered as Evidence
Location evidence comes from satellite positioning, from network cell sites, from wireless network observations and from application check-ins, each with a different accuracy. All of it places a device rather than a person. Interpreting it responsibly means establishing which method produced each point and what margin that method carries.
Social Media Posts and Their Custody
Social media evidence presents two distinct problems. The first is capture: a post can be edited or removed at any time, so a record has to be made that will outlive it. The second is attribution: an account is not a person, and connecting a post to an individual requires evidence beyond the display name on the profile.