What 911 Call Analysis Is (and Why It's Still in Courtrooms)

Your client's fate may hinge on how they sounded on a 911 call. That's the core problem with 911 call analysis: a technique that claims to detect guilt from a caller's word choice, phrasing, and emotional tone is still being used by police and prosecutors to argue culpability, even though study after study has debunked its reliability.

The method traces back to Detective Tracy Harpster's 2006 master's thesis built on just 100 calls, research the Innocence Project notes was intended as exploratory only. The current version, the 911 COPS Scale, has grown to 38 "guilt indicators" across six categories, from how the call is made to resistance to the dispatcher. At least eight replication studies have failed to validate those indicators.

Yet the technique has real courtroom reach. ProPublica documented more than 100 cases across 26 states where law enforcement used Harpster's methods, and Harpster claims the analysis helped solve over 1,500 homicides. Departments pay $3,500 for eight hours of training, often with taxpayer money, then officers testify about what the caller's words "really" meant.

Here's the structural problem: courts have repeatedly rejected this as expert testimony. So prosecutors route it through lay testimony instead. Officers and dispatchers who attended a training describe "guilt indicators" without ever proffering the underlying method, which lets the technique largely escape Daubert or Frye scrutiny. Cornell Law Review calls it "shadow forensics," placing it alongside debunked bitemark and hair microscopy evidence.

For defenders, recognition is the first step. Watch for these markers in discovery and at trial:

• Officer testimony interpreting the caller's phrasing, pauses, or "passive defenses" as signs of deception
• A detective or dispatcher who attended 911 call analysis training but testifies without expert qualification
• Arguments that the caller "did not seem emotional enough," the bias Dr. Jessica Salerno's research shows drives suspicion regardless of actual guilt
• Prosecution walkthroughs of the call audio and transcript designed to "create lasting impressions" for the jury

The stakes are not hypothetical. Huwe Burton served more than 20 years before his 2019 exoneration, convicted in part because an officer testified he lacked emotional display. Robert Roberson, later diagnosed with autism spectrum disorder, has spent over two decades on Texas death row after his flat affect was read as guilt.

If your case involves call evidence, start with your own grounded transcript rather than relying on the state's characterization. A diarized, word-timestamped transcript lets you pin down each claimed indicator to the exact second of audio, which is the workflow behind Evidence Copilot's 911 call transcript analysis for defense teams. The same grounding matters for custodial calls, where jail call transcription gives you citable text instead of an officer's paraphrase.

The Shadow Forensics Problem: How Prosecutors Sneak It Past Daubert

The most dangerous thing about 911 call analysis isn't the science. It's the delivery route. Courts keep rejecting it as expert testimony, so prosecutors have stopped offering it as expert testimony. Instead, a trained detective or dispatcher takes the stand and describes "guilt indicators" in your client's call, and the whole exercise slips past Daubert without a hearing.

Christopher Lau of the Wisconsin Innocence Project calls this the "shadow forensics" problem in the Cornell Law Review: the underlying methods are never disclosed, the founders are never proffered as experts, and the technique "largely escapes Daubert or Frye evidentiary challenge" because it arrives dressed as ordinary lay observation. He situates it alongside bitemark evidence and hair microscopy, junk disciplines that have contributed to hundreds of exonerations.

The evasion isn't accidental. It's strategy, taught at the source. University of Cincinnati Law Review analysis documents that Harpster himself provides ten tips for prosecutors to use the method without ever seeking expert qualification, and that prosecutors openly acknowledge the method "will not hold up to scientific scrutiny." The Innocence Project reports the same playbook: introduce it indirectly through a trained detective or dispatcher rather than through the analyst. ProPublica has now documented more than 100 cases across 26 states where law enforcement used Harpster's methods.

Why does this work? Because FRE 701 lets a lay witness give opinions based on rational perception, and a jury hears "I noticed the caller seemed calm" as common sense rather than as a $3,500 training product with 38 checklist indicators. The fix is forcing the 701/702 question: if the officer's opinion rests on specialized knowledge from a Harpster course, it's expert testimony, and it must survive Daubert or Frye. That's where the method collapses, because:

• The foundational study was 100 calls, intended as exploratory research only
• At least eight replication studies have failed to validate it
• Harpster and co-author Dr. Susan Adams have never released their data or methodology
• Courts in Alabama, Nevada, and Michigan have already rejected the analysis when it was offered as expert testimony

Your battleground is the lay/expert line, and you should fight it at the earliest opportunity: preliminary hearing, motion in limine, voir dire of the testifying officer. Ask what trainings they've attended, what materials they received, and whether they were taught to identify "guilty indicators." Once that's on the record, the 701 safe harbor evaporates.

Practically, you also need your own grounded transcript of the call, because the prosecution's playbook is to guide juries through audio and transcript together to "create lasting impressions." Services like Evidence Copilot's 911 call transcript analysis give you timestamped, diarized text so you can pin each claimed indicator to the actual moment in the call, and jail call transcription covers the same need when the state pivots to recorded calls from custody. Every indicator should be tested against the call itself, never against the officer's characterization of it.

Your Objection Playbook: FRE 701/702 and the Parallel Hearsay Challenge

The moment a dispatcher starts explaining what a caller's word choice "really means," you have two objections ready, and they attack on different axes. Run them together.

Axis one: the FRE 701/702 objection. When an officer's testimony about "guilt indicators" rests on specialized knowledge from a $3,500 training, it's expert testimony that hasn't been qualified. Prosecutors know this; that's why, as the Cornell Law Review documents, they route 911 call analysis through lay witnesses so it "largely escapes Daubert or Frye evidentiary challenge." Object under FRE 702 and demand qualification. Then build the Daubert record from the method's opacity:

• No testability: Harpster and co-author Dr. Susan Adams have never released their underlying data or explained how they obtained their sample calls, per the University of Cincinnati Law Review.
• No peer review or replication: at least eight replication studies have failed to validate the indicators, and the foundational study was a 2006 master's thesis built on just 100 calls.
• No error rate: without released data, no one can compute one. That's a Daubert factor failure on its face.

Courts have already rejected the technique when proffered as expert testimony, including in Alabama, Nevada, and a Michigan murder conviction that was overturned. Your job is to force the state to defend the method in daylight rather than let it ride in through a lay witness.

Axis two: the parallel hearsay objection. Separate from call analysis, officer testimony repeating a caller's assertions to prove their truth is inadmissible hearsay. In U.S. v Nelson (6th Cir. 2013), a conviction was reversed on exactly this ground: the caller's suspect description was hearsay, and the error wasn't harmless. The fallback demand matters here. If the state insists the call explains officer conduct, ask for the background-only formulation: "police received a 911 call," with no details. That preserves the legitimate purpose while stripping the prejudicial substance, and it sets up your harmless-error argument on appeal if the court overrules you.

Both objections get stronger when you can quote the call itself. A diarized, word-timestamped transcript of the 911 audio (or the related jail call transcription, if recorded calls are in discovery) lets you confront each claimed indicator against what the caller actually said, second by second, instead of against the officer's characterization. Evidence Copilot's 911 call transcript analysis exists for exactly this workflow: cited moments, human review, no invented conclusions.

The Emotional-Display Bias: Your Cross-Exam Counter-Narrative

When an officer tells a jury your client "did not seem emotional enough" on the 911 call, that observation is doing more work than any fingerprint. It is also, according to peer-reviewed research, exactly the kind of judgment that convicts innocent people.

Dr. Jessica Salerno's studies found that both civilians and police officers become significantly more suspicious of callers who express low emotion. The finding held in follow-up studies using real 911 calls with known outcomes: the association between flat emotional display and perceived guilt emerged regardless of whether the caller was later convicted or exonerated. Officers were the harshest audience, requiring callers to sound highly emotional before their suspicion eased. As Salerno puts it, the system "continues to rely heavily on subjective assumptions about what grief, fear, and innocence should look like."

Two cases show the stakes. Huwe Burton served more than 20 years before his 2019 exoneration, convicted at 16 largely because an officer testified he "did not seem emotional enough." Robert Roberson, later diagnosed with autism spectrum disorder, has spent more than two decades on Texas death row after his flat affect was read as guilt, narrowly avoiding execution three times. Both stories are documented in the Innocence Project's analysis of 911 call analysis.

Use this at cross like this:

• Establish the officer attended the $3,500, eight-hour training, then walk them through Salerno's finding: low emotion triggers suspicion whether the caller is guilty or innocent, so the "indicator" cannot distinguish between the two.
• Ask whether the officer asked about trauma, shock, autism, or any other explanation for flat affect before forming a suspicion. The answer is almost always no.
• Confront with Burton and Roberson by name. "You knew, didn't you, that people have served decades based on this exact assumption?"
• Offer the counter-narrative affirmatively: neurodivergent flat affect, acute trauma response, and dispatcher-driven call structure all produce the same low-emotion display the state calls a guilt indicator.

In motions, pair the research with the FRE 701/702 fight. If the officer's "he wasn't emotional enough" opinion rests on specialized knowledge from the training, it is expert testimony that never qualified, and the documented court rejections of Harpster as an expert give you the record. Salerno's own recommended remedy is blunt: defense attorneys should learn to identify and challenge attempts to introduce call analysis through lay testimony.

Practically, this means you need the call itself, not the officer's characterization of it. Get a diarized, word-timestamped transcript early so you can cite the exact second of each claimed indicator and test it against the audio. A 911 call transcript analysis workflow built for defense teams, with every output tied to the source moment and marked for human review, makes that feasible before the preliminary hearing. The same applies to jail call transcription, where tone-based inferences show up just as often. The emotional-display research gives you the science; the transcript gives you the ground truth to cross on.

Ground the Audio Early: Building Your Own 911 Call Analysis Workflow

The prosecution's playbook is to walk juries through the audio and transcript to "create lasting impressions of this critical evidence". You can't counter that with a police report summary and a memory of the call. You need your own transcript, grounded to the second, before the preliminary hearing.

Start by getting a timestamped, diarized transcript of the 911 call the moment it lands in discovery. Word-level timestamps matter because "guilt indicators" are word-by-word claims: the COPS Scale lists 38 indicators across six categories, and each one points at a specific phrase or pause. If you can't cite the second, you can't impeach the characterization.

Then run every claimed indicator through a simple audit trail:

• Play the exact moment. The officer says the caller used "passive defenses." Find the phrase in the transcript, play those three seconds, and write down what you actually hear.
• Separate observation from training. Note which claims are perceptible to any listener and which require the $3,500 eight-hour course to see. That split is your FRE 701/702 objection.
• Cross-reference the police report. Line up the call against the report's narrative. Inconsistencies between what the caller said and what the officer wrote are impeachment material independent of call analysis.
• Flag emotional-display claims. Research shows suspicion of "not emotional enough" callers tracks perceived guilt regardless of whether the caller was later convicted or exonerated. Anything phrased as affect judgment goes on this list.

Evidence Copilot's 911 call transcript analysis and jail call transcription are built for exactly this loop: diarized, word-timestamped transcripts, every AI answer cited to the exact second, every output marked pending human review. Auto-signals compare the call against the police report on ingest and surface candidate inconsistencies before anyone asks.

The institutional ground is shifting under the technique. North Carolina's Office of Indigent Defense Services issued a statewide warning to attorneys, Fair and Just Prosecution called on member prosecutors to review past cases, and the Illinois Supreme Court agreed to revisit Jessica Logan's conviction. Cite those in your motions. Courts notice when 26 states' worth of use is under review.

If you want to see the workflow on a real call, book a 20-minute walkthrough. Bring the audio; we'll ground it live.