What 911 Call Analysis Actually Is (and Why It Matters to Your Case)

A caller says "hi" before reporting her baby stopped breathing, and a detective flags the greeting as evidence of guilt. That's 911 call analysis in a nutshell: a technique that claims to identify deception from word choice, cadence, and linguistic cues in emergency calls. If you defend cases where someone dialed 911 and later became the suspect, you need to know how this works, because it's probably already in your discovery.

The technique comes from retired Deputy Chief Tracy Harpster, whose COPS Scale lists 38 indicators that supposedly distinguish guilty from innocent callers. His original research rested on just 100 calls, half from guilty callers and half from innocent ones, and a later book expanded the sample to 200. Harpster has consulted on at least 2,500 homicide investigations as of March 2024, according to the University of Cincinnati Law Review, and ProPublica's investigation documented more than 100 cases across 26 states where police applied his methods.

Here's the part that should stop you cold: Harpster has never testified as an expert in a criminal trial. He was rejected as an expert in an Alabama courtroom back in 2009. His underlying data has never been publicly released, and a 2020 independent attempt to apply the methods, per the Chicago Criminal Lawyer Blog, failed to reproduce usable results. Yet convictions still stand on this foundation, including a Michigan murder conviction that was later overturned precisely because 911 call analysis shouldn't have come in.

So how does an unproven technique keep landing convictions? Through a workaround. Harpster advises prosecutors to have witnesses "guide juries through both the audio and transcript of the calls" without ever seeking expert qualification, which smuggles the analysis past Daubert and Frye scrutiny. Collin Derrig put it bluntly: the work has been "mixed into in-court statements and lay testimonies to sneak it past the standards for expert testimony." That means the danger isn't a disclosed expert report. It's a detective's lay opinion about your client's "guilt indicators."

This is why 911 call analysis screening belongs in every defender's discovery review, not just homicide specialists' checklists:

• Ask the prosecutor directly whether call analysis shaped the investigation, interrogation, or charging decision.
• Search discovery for correspondence with Harpster or his trainees, training records, or any mention of the "COPS Scale."
• Watch for officers offering lay opinions about "staged" behavior or deception cues in the call, and object under FRE 701.
• Check whether the interrogation was built on call-analysis handouts, as in a 2012 Ohio case where a detective's handout preceded a confession.

The practical first step is getting the call itself into a workable transcript, since the analysis lives and dies in the words. A 911 call transcript analysis workflow with word-level timestamps lets you compare what was actually said against what the detective claims it meant, and the same discipline applies when you pair it with jail call transcription for the rest of the call record. Evidence Copilot built its review workspace around exactly this kind of quote-level comparison, so every challenge you raise points to a specific second of audio.

The technique spread quietly for two decades, and the screening burden falls on the defense. North Carolina's indigent defense office has now warned attorneys statewide, and the Illinois Supreme Court agreed to re-examine a case built on the method. The question in your case file isn't whether 911 call analysis is valid. Courts are increasingly answering that one for you. The question is whether it touched your case without anyone telling you.

The Daubert and Frye Problems With 911 Call Analysis

If a prosecutor tries to put 911 call analysis in front of your jury as science, you have the strongest admissibility record in recent memory to fight it. The technique has never survived a Daubert or Frye challenge in any reported case where its developer, Tracy Harpster, testified, because he has never qualified as an expert at trial at all. An Alabama court rejected him as an expert witness back in 2009, according to the University of Cincinnati Law Review.

Walk the five Daubert factors against the actual record and the technique fails every one. Consider what the underlying research consists of: a thesis built on 100 calls, a 2017 book expanding to 200 calls, and a 38-indicator "COPS Scale." Harpster and his co-author have never released how they obtained those calls or made the data public.

Here's how each factor breaks down:

• Testability: A 2020 independent study attempted to apply the methods and "failed miserably to reproduce actionable results," per reporting on the technique. The one real-world test we know of came back negative.
• Peer review and publication: No published replication exists. The data behind the COPS Scale has never been made available for scrutiny.
• Error rate: There is no known error rate. Nobody, including Harpster, has measured one.
• Controlling standards: The 38 indicators are subjective interpretive cues, such as a caller saying "hi" being flagged as suspicious politeness in a documented 2018 Washington case.
• General acceptance: The NACDL's president called the practice "dangerous and insidious," and the Colorado Bureau of Investigation says it should never be a deciding factor in an arrest or conviction, per ProPublica's investigation.

If you're in a Frye state, the argument is even simpler: no methodology that has never been replicated, never had its data released, and whose originator has never passed voir dire qualifies as "generally accepted" in any relevant scientific community. The Michigan v. Spitler reversal, where an appellate court found 911 call analysis should never have been admitted, is your roadmap.

The real fight in 2026 isn't the expert witness. It's the workaround. Because Harpster can't survive expert qualification, he advises prosecutors to "guide juries through both the audio and transcript of the calls" to create "lasting impressions" without seeking expert status. That's 911 call analysis smuggled in through lay testimony, and it violates FRE 701, which bars lay opinions based on specialized knowledge.

When a detective testifies about a caller's "guilt indicators," object immediately: the opinion isn't rationally based on perception, it's applied methodology dressed up as intuition. Demand the prosecution's discovery for any Harpster correspondence or training records, then build your impeachment record against the officer's claimed expertise.

For building that record, a grounded 911 call transcript analysis workflow with timestamped citations into the exact seconds of audio lets you quote the call against the detective's testimony line by line. The same applies to jail call transcription when the state leans on recorded calls elsewhere in its case.

How Prosecutors Sneak It Past the Expert Witness Standard

Tracy Harpster has never qualified as an expert witness in a criminal trial. That's not because prosecutors stopped using his method. It's because he told them not to put him on the stand. Instead, he advises prosecutors to "guide juries through both the audio and transcript of the calls," letting a trained detective walk the jury through the call's "guilt indicators" as if the detective were just sharing personal observations. As Collin Derrig wrote in the University of Cincinnati Law Review, the analysis "has been mixed into in-court statements and lay testimonies to sneak it past the standards for expert testimony."

The result is a lay testimony workaround. A detective who completed Harpster's training takes the stand as a fact witness, plays the 911 call, and offers opinions that sound like common sense but map directly onto the 38-indicator COPS Scale. No Daubert hearing. No Frye hearing. No error rate, no peer review, no cross-examination of the underlying methodology, which a 2020 independent study already "failed miserably to reproduce."

Watch for this language in the transcript. It's the tell that the witness is applying specialized methodology while dressed as a lay observer:

• Characterizing the caller's word choice as a "guilt indicator" or "deception cue"
• Describing the call as "staged," "too polite," or lacking "excitement" consistent with a real emergency
• Commenting on the order of information in the call, such as asking about the suspect's description before the victim's condition
• Testifying that the caller "didn't want the call traced" or showed "avoidance" in specific phrasing

The Washington case is the clearest example of how far this stretches. A detective flagged a young mother's first word, "hi," as suspicious because being too polite could indicate she was hiding something. Three pathologists couldn't determine whether the baby's death was a homicide. The call analysis got the death reclassified anyway, and the mother took an Alford plea rather than risk life in prison.

Your objection framework is FRE 701. Lay opinion is limited to what is rationally based on the witness's perception, helpful to the trier of fact, and not based on scientific, technical, or specialized knowledge. A detective interpreting a 911 call through Harpster's trained indicators is doing exactly what FRE 702 exists to screen: applying a specialized methodology to draw conclusions about guilt. The objection is that the testimony's foundation isn't the detective's perception of the call. It's the COPS Scale, and that methodology has never survived admissibility review anywhere Harpster himself has testified.

Ask for the detective's training records in discovery. If Harpster's name or the COPS Scale appears, you have your foundation for the motion in limine, and you know precisely which transcript passages to flag before trial.

One practical note: catching these opinions requires a transcript you can search line by line. When we review 911 audio, we work from a diarized, timestamped transcript so every "guilt indicator" opinion can be tied to the exact words the detective relied on, the same way we handle jail call transcription for cross-examination prep.

Using 911 Calls Affirmatively in Your Defense

The same recording the prosecution wants to weaponize can be your best exhibit. A 911 call is a contemporaneous, unfiltered snapshot of the scene, captured before anyone had a story to stick to, and it often contradicts what the police report says later.

The procedural path is straightforward. Per standard practice guidance, you take three steps: list the call as evidence with the prosecutor, subpoena the right witness, and get a ruling from the judge.

• List the 911 call on your exhibit list and disclose it to the prosecutor.
• Subpoena the custodian of records who maintains 911 recordings in that jurisdiction (or the call-taker herself).
• Request a judicial ruling on whether the actual recording may be played at trial versus a transcript read to the jury. That decision belongs to the judge, not the parties.

One caution before you hit play. The same guidance flags that playing the recording may waive certain rights against self-incrimination if your client is the caller. If your client's voice is on that line, think hard about whether the impeachment value outweighs opening that door.

The payoff is real. The call-taker's standardized questions, location, callback number, nature of the emergency, per 911.gov, produce a structured record you can line up against the police narrative. When the report says your client was "calm and uncooperative" but the audio captures a panicked caller performing CPR instructions, the jury hears the truth. The Washington case in the research is the mirror image: a detective flagged the mother's first word, "hi," as a guilt indicator, and that reading only survived because the raw context was reframed. The audio itself is the antidote.

Practically, you need the call transcribed with word-level timestamps so you can cite the exact second in a motion or cross. Tools like Evidence Copilot handle 911 call transcript analysis alongside the rest of discovery, and the same workflow covers jail call transcription when those recordings matter too. Given that Harpster has consulted on at least 2,500 homicide investigations as of March 2024, per the University of Cincinnati Law Review, the call in your case may already have been analyzed by the other side. You should know it just as well.

Discovery Checklist and Workflow for Reviewing 911 Evidence

Discovery moves fast, and 911 call analysis evidence hides in places most defense teams don't think to look. A screening pass on day one saves you from missing a Harpster-trained detective who shaped the entire investigation.

Start with a direct ask to the prosecutor. Request the full 911 audio, CAD records, call-taker notes, and any correspondence with outside consultants. Specifically ask whether any investigator applied "statement analysis" or "behavioral analysis" to the call, since defense scholarship documents how prosecutors embed the technique into lay testimony to dodge Daubert review.

Then mine the discovery dump for three specific artifacts:

• Training records showing the detective completed Harpster's 911 call analysis course or holds a certificate referencing the COPS Scale (the 38-indicator rubric Harpster and Adams developed from a 100-call thesis sample)
• Email or memo correspondence with Tracy Harpster or a known trainee, including "consultation" notes and handouts left in the interrogation room
• Police report language that scores the caller's word choice: "failed to show concern," "inappropriate affect," "staged call," or "guilt indicators"
• Interrogation footage where the detective references the 911 call as proof of deception before the suspect gives a statement
• Jail calls where your client re-explains the same events, which lets you compare spontaneous phrasing against the original 911 phrasing

Review the 911 audio alongside bodycam and interrogation footage in one workspace, not three windows. You want word-level timestamps on the call so you can quote the exact second a phrase lands, then pull the matching bodycam moment where an officer characterizes the same event differently. 911 call transcript analysis paired with 911 call transcript analysis tooling gives you diarized, citable text you can drop straight into a motion or cross-exam outline.

Jail calls deserve the same treatment. Running jail call transcription alongside the 911 transcript surfaces phrasing drift between the spontaneous call and later retellings, which cuts both ways and you should know which way before trial.

Log every hit in a running review memo with file name, timestamp, and a one-line note on why it matters. When you file your Daubert or Frye motion, that memo becomes your factual appendix. Screening for these artifacts early is the difference between a suppression fight you saw coming and a surprise at the preliminary hearing.

Evidence Copilot flags candidate inconsistencies between the 911 audio, police report, and bodycam automatically on ingest, so the Harpster-trained detective's report-vs-call gap shows up before you ask. Book a 20-minute walkthrough at evidencecopilot.com/book if you want to see the workflow on a real case file.