What 911 Call Analysis Is and Why It's in Your Discovery

The 911 call comes in at 2:14 a.m., and by the time your client is charged, a detective somewhere has already scored how the caller asked for help. That scoring has a name, a training program, and a body count of convictions behind it, and you need to understand exactly what it is before you can attack it.

911 call analysis, sometimes called the 911 COPS Scale, is a technique developed by Tracy Harpster that teaches law enforcement to read guilt and deception from a caller's word choice, cadence, and grammar. The FBI and other agencies helped it spread, and ProPublica documented more than 100 cases across 26 states where departments used it. Harpster himself has consulted on at least 2,500 homicide investigations, according to reporting on the technique's reach.

Here's the part that matters for your motion practice. The technique rarely shows up as expert testimony. Instead, officers and 911 operators who sat through the training testify as lay witnesses about what they heard, which means the underlying methodology never gets tested under Daubert or Frye. Cornell Law Review calls this "shadow forensics": forensic techniques shielded from evidentiary oversight because the founders are never proffered and the data is never disclosed.

The published science doesn't hold up under pressure, either. In a study of 175 calls, a panel of officers and a forensic psychologist agreed on whether voice modulation was present only 72% of the time. The same FBI research found that over 80% of both innocent and guilty callers used the word "just," and the authors concluded these markers "should not be used on their own for evaluating veracity versus deception."

• The technique has no disclosed, peer-reviewed validation, and its creators have never released their underlying data.
• Adoption continues anyway: departments and prosecutor offices keep paying for training and keep using it to seek convictions.
• The backlash is real. NACDL, the Innocence Project, and Fair and Just Prosecution are pushing bans and reviewing past convictions.
• Some agencies now quietly recast the method as an "investigative lead" rather than evidence, which is an admission worth quoting.

So the core problem is this: an unvalidated technique arrives in your discovery dressed as evidence, usually through the mouth of an officer rather than a scientist. Beating it starts with the transcript itself. A disciplined 911 call transcript analysis lets you line up the caller's actual words against the police report, the bodycam, and the later interviews, and the gaps you find are where confirmation bias shows. Pair that with jail call transcription and you can track how the narrative hardened between the 2:14 a.m. call and the interrogation room.

If you're building that comparison by hand across a full discovery dump, Evidence Copilot was built for exactly this: upload the 911 audio, the report, and the bodycam, and every answer it gives you cites the exact second or page it came from, with all AI output marked for your review. You can book a 20-minute walkthrough or start on the Solo plan for $39 a month when you're ready.

The Science Doesn't Hold: FBI Data on Voice Modulation and Minimization

When a prosecutor's expert tells the jury that a caller's voice "modulated" or that the word "just" proves minimization, the cross-examination writes itself, because the government's own research says these markers don't work. The FBI's Law Enforcement Bulletin, hardly a defense-friendly outlet, published data that guts the two most common pillars of 911 call analysis.

Start with voice modulation. In a study of 175 calls, a panel of law enforcement officers and a forensic psychologist could agree on whether voice modulation was even present in only 72% of cases, according to the FBI's own analysis. If trained raters cannot agree on whether the phenomenon exists in a given call, it cannot reliably distinguish guilty callers from innocent ones.

The literature is worse than merely weak; it points in opposite directions. An early study found that callers with limited change in tone were more likely to be guilty. A follow-up by FBI authors found the opposite: offenders were more likely to exhibit voice modulation. And when researchers compared staged homicides to true suicides, voice modulation failed to distinguish between the groups at all. That is not a contested edge case. That is a foundational indicator flipping its meaning between studies.

Minimization fares no better. The claim, popularized by technique architect Tracy Harpster's training program, is that guilty callers soften their language with words like "just." The empirical record:

• Over 80% of both innocent and guilty callers used the word "just," meaning its presence tells you almost nothing.
• Involved callers averaged 4 uses per call versus 2.5 for uninvolved callers, a frequency gap too small to sort individuals in a single case.
• 52% of homicide offenders gave their own location before discovering the body, compared to 36% of uninvolved callers, the kind of overlapping distribution that produces false positives in both directions.

The FBI authors' conclusion is the sentence you want on the screen during cross: voice modulation and minimization "should not be used on their own for evaluating veracity versus deception." They also warned that investigators analyzing minimization may fall into confirmation bias, which is exactly the framing you want when the officer who "analyzed" the call was also the one who built the case.

So when the state's witness testifies that your client "minimized," put the numbers to the jury. Ask the officer whether they know that 8 out of 10 innocent callers say "just" too. Ask whether they can cite a single validated study supporting the 911 COPS Scale, given that its creators have never released their data publicly, as noted in a University of Cincinnati Law Review analysis.

Statistical ammunition only works if it's anchored to the actual audio. A grounded transcript review, timestamped and diarized, lets you put the caller's exact words next to the expert's characterization and show the gap. Tools built for that workflow, including Evidence Copilot's 911 call transcript analysis, keep every quote tied to the second of audio it came from, so the impeachment survives scrutiny. The same discipline applies when you compare pre-arrest 911 statements against post-arrest jail calls through careful jail call transcription, because selective interpretation doesn't stop at the 911 line.

Attacking Admissibility Through the Shadow Forensics Framework

Prosecutors have been getting 911 call analysis testimony in front of juries for years, and defense attorneys have largely been unable to stop it. The reason isn't that the science holds up. It's that the science never gets tested.

Christopher Lau's Cornell Law Review analysis calls this "shadow forensics": forensic techniques that escape Daubert and Frye review because the methodology is never formally proffered as expert testimony. Instead, a trained 911 operator or detective takes the stand and testifies about "indicators of guilt" they observed on the call. The creator of the 911 COPS Scale, Tracy Harpster, never testifies. The underlying dataset is never produced. The court never holds a reliability hearing. The jury hears the conclusion without ever seeing the foundation.

This is the structural defect you attack.

Most defense motions to exclude 911 call analysis target the wrong layer. They try to exclude the technique as unreliable under Frye or Daubert. That motion often fails because the prosecution isn't offering the technique as expert testimony. They're offering a lay witness who has been "trained" to recognize certain speech patterns. The witness becomes the container, and the methodology hides inside.

To crack this open, you have to reframe the challenge around what the witness actually did. If a detective applied the COPS Scale to score a caller's language, voice modulation, and word choice, that is forensic analysis. It doesn't matter that the detective isn't called an "expert." The methodology is being applied, and the methodology has never been peer-reviewed, never validated against innocent callers at scale, and never released for scrutiny.

The motion strategy Lau outlines has three pressure points. File them together:

• Demand methodology disclosure. Request the complete 911 COPS Scale, including all scoring criteria, training materials, and the underlying validation dataset. The data has never been released publicly, which is itself a Daubert problem.
• Force the creator onto the stand. Move to require the State to produce the person who developed and validated the technique. If Harpster or his designees trained the witness, the training materials and validation studies come with them.
• Attack the indicators one by one. The FBI's own Law Enforcement Bulletin research found that voice modulation and minimization cannot reliably distinguish guilty from innocent callers. Over 80% of both groups use the word "just." Expert agreement on whether modulation is even present sat at 72% across 175 calls. Cite the FBI against the prosecution.
• Argue shadow forensics in limine. Frame the motion as a due-process challenge: the court cannot admit forensic conclusions without the underlying science, regardless of who delivers the testimony.

The ProPublica investigation documented more than 100 cases across 26 states where this method was deployed, and the NACDL has now classified it alongside bitemark analysis as forensic junk. That institutional weight matters when you're asking a judge to do something courts have been reluctant to do: treat a cop's testimony as expert opinion and subject it to the gatekeeping it has always deserved.

When you're building the factual record to support the motion, the work is in the transcripts. A timestamped, diarized 911 call transcript analysis lets you show the court exactly what the caller said, what the officer later claimed they heard, and where those two things diverge. That gap is where the shadow forensics collapses.

Building a Transcript-Driven Counter-Narrative

The 911 call is often the first version of events anyone hears, and by the time it reaches a jury it has been filtered through a police report, a training bulletin, and a theory of the case. Your job is to get back to what the caller actually said, at the second they said it.

Start by pulling the raw audio, not a summary. Then build a timestamped, diarized transcript so every line is tied to a speaker and a moment you can cite. From there, the real work is comparison. FBI researchers recommend that deception detection is more fruitful when statements in the call are checked against case facts, written statements, and interviews, rather than judged on vocal cues in isolation (FBI Law Enforcement Bulletin).

Line the call up against four other records:

• The police report, to see what the officer wrote down and what got left out.
• Bodycam and dashcam, to test whether the report matches what was said on scene.
• Any later interview or interrogation, to track how the account changed.
• Jail calls, to compare pre-arrest statements with post-arrest conversations.

The numbers matter here. In a study of 175 calls, a panel of officers and a forensic psychologist agreed on whether voice modulation was even present only 72% of the time (FBI Law Enforcement Bulletin). Over 80% of both innocent and guilty callers used the word "just," which makes minimization arguments far weaker than prosecutors suggest (FBI Law Enforcement Bulletin).

This is where confirmation bias hides. An investigator who believes the caller is guilty reads the same words as evidence of guilt. A timestamped transcript lets you show the sequence plainly: what the caller said first, what the officer wrote later, and where the two drift apart. For defense teams, a 911 call transcript analysis workflow keeps those comparisons grounded in citable source moments instead of memory, and jail call transcription extends the same discipline to post-arrest audio. Every output is AI-generated and pending your review, with originals left untouched.

If you want to see how this runs on a real discovery dump, book a 20-minute walkthrough at evidencecopilot.com/book.

Post-Conviction Review and Staying Current in 2026

The same "shadow forensics" problem that makes 911 call analysis dangerous at trial creates a parallel crisis in the appellate pipeline. ProPublica documented more than 100 cases across 26 states where the method was used to identify suspects, and a growing number of those convictions are now under formal review.

The most visible test case sits in Illinois. The state Supreme Court agreed to hear People v. Jessica Logan, a conviction built in part on 911 call analysis testimony. Defense counsel Josh Tepfer, who has helped exonerate 288 wrongfully convicted people, is arguing that the technique's lack of empirical support makes its admission a violation of due process. A ruling against the State would reopen dozens of pending cases nationwide.

Two institutional actors are doing the heavy lifting behind the scenes. The Innocence Project is screening convictions where the method featured at trial, and Fair and Just Prosecution has called on elected prosecutors to initiate post-conviction review in their own jurisdictions. NACDL President Nellie King framed the moment bluntly: the technique belongs on the "junk pile of fake scientific theories." That posture matters because it gives defense filings a recognized institutional backer rather than a lone expert.

For practitioners with existing clients, the practical move is to pull the trial transcript, confirm whether a 911 call analysis opinion was admitted, and check the conviction date against the state's post-conviction statute of limitations. The strongest argument is structural: the technique's own creator has never released the underlying dataset, so the defense never had a meaningful chance to test it under Daubert or Frye.

Staying current in 2026 means watching three signals. Track state-court rulings and AG opinions as they drop, since bans are arriving piecemeal rather than federally. Read the NACDL's May 2026 guidance on forensic junk science for the framing language courts are accepting. And keep an eye on the institutional shifts, because when prosecutors' own associations call for review, the political cover for new motions grows fast.

For defense teams handling active dockets, the workflow is straightforward. Run 911 call transcript analysis to compare the caller's statements against the police report and any bodycam, then layer in jail call transcription to see whether the post-arrest narrative matches what the caller actually said at the scene. Every grounded citation you build now is one you can repurpose later if the conviction gets revisited.