What 911 Call Analysis Is and Why It's Showing Up in Your Cases

Somewhere in your discovery dump, there may be a piece of junk science you've never heard of, and it won't announce itself. It arrives disguised as a detective's "training and experience" or a dispatcher's impression of how your client "sounded" on the 911 call. By the time you recognize it, the framing has already shaped the investigation.

911 call analysis is the practice of inferring guilt or innocence from how a caller speaks: word choice, cadence, grammar, emotional tone. The best-known version is the "911 COPS Scale©" developed by former Ohio homicide detective Tracy Harpster, who claims to have reviewed more than 2,500 equivocal death cases over roughly two decades, according to Governing's reporting. The foundational study behind the method examined only 100 calls and was intended as exploratory research, and at least eight replication studies have failed to validate it, per the Innocence Project.

Despite that record, it's everywhere. ProPublica's investigation documented more than 100 cases across 26 states where law enforcement used Harpster's methods, and police departments have paid up to $3,500, often taxpayer-funded, for roughly eight hours of training in the technique.

The institutional backlash is now nationwide:

• Prosecutors, judges, and defense organizations have called for banning the technique, reviewing past convictions, and sanctioning prosecutors who introduced it knowing it was inadmissible.
• North Carolina's Office of Indigent Defense Services issued a statewide warning telling attorneys to watch for appearances of 911 call analysis in their cases.
• Fair and Just Prosecution called on its member prosecutors to review past cases for its use.
• The Illinois Supreme Court agreed to review People v. Logan, a conviction built partly on a Harpster-trained detective's testimony, after ProPublica's reporting.

NACDL president Nellie King put it bluntly: the criminal legal system can add 911 call analysis to "the junk pile of fake scientific theories."

Here's the problem for your practice. Because the method can't survive scientific scrutiny, prosecutors often smuggle it in through lay testimony from trained detectives and dispatchers, so jurors hear "guilty behavior" inferences without ever learning the disputed methodology. That means it may already be sitting in your discovery, buried in a police report's characterization of the call or an officer's anticipated testimony, and you won't see it unless you know what to look for.

The first step is getting the actual call in front of you, word for word. A timestamped 911 call transcript analysis lets you confront the officer's characterization against what your client actually said and when. The same discipline applies to recorded calls from custody, where tone-based inferences creep in too; jail call transcription with word-level timestamps gives you the record to impeach with. Evidence Copilot built its review workflow around exactly this: every claim about the audio should trace back to a specific second you can play in court.

The Lay-Testimony Workaround: How Call Analysis Smuggles Past Daubert

When prosecutors can't get 911 call analysis past Daubert, they don't drop it. They just stop naming it. Instead of an expert describing the "911 COPS Scale," a trained detective takes the stand as a lay witness and testifies that the caller "didn't sound right," that her answers were "evasive," that he "did not seem emotional enough." Jurors hear the guilty-behavior inference without ever learning it comes from a disputed methodology.

That's the workaround ProPublica's investigation documented: prosecutors presenting call analysis indirectly through trained detectives and dispatchers, with some conceding the method "will not hold up to scientific scrutiny" and getting "creative" instead. The Innocence Project reports at least eight replication studies failed to validate the method, so the lay-testimony route is often the only path left.

Dr. Jessica Salerno's research explains why this works so well on juries. Both civilians and police become significantly more suspicious of low-emotion callers, and for police officers, even moderate emotion wasn't enough to dispel suspicion. Her follow-up studies with real 911 calls found the pattern held regardless of whether the caller was later convicted or exonerated. As she put it, "Mistaken suspicion is rarely corrected by the investigation that follows. More often, it shapes the investigation itself."

The case history shows what that bias costs:

• Huwe Burton, 16 years old, called 911 after finding his mother murdered. An officer testified he "did not seem emotional enough." He served more than 20 years before his 2019 exoneration.
• Robert Roberson spent 20+ years on Texas death row after his flat affect was read as guilt. He was later diagnosed with autism spectrum disorder.
• Jessica Logan was convicted after a Harpster-trained detective testified about her 911 call. The Illinois Supreme Court agreed to review the case after ProPublica's reporting.

So how do you catch it in your own discovery? Watch for the telltale characterizations: a detective's report describing the client's 911 demeanor, anticipated testimony about whether the caller "sounded" consistent with a genuine emergency, dispatchers opining on tone. North Carolina's Office of Indigent Defense Services warned attorneys statewide to watch for exactly these appearances. If the witness has taken Harpster's training (departments pay up to $3,500 for roughly 8 hours of it), that's your Daubert argument hiding in a lay-opinion wrapper.

The counter is the call itself. A word-level timestamped transcript lets you confront the officer's "he didn't sound upset" characterization against the actual audio, second by second, on cross. That's the workflow we built 911 call transcript analysis around at Evidence Copilot: diarized, timestamped transcripts with citations into the exact moment, so the officer's characterization meets the record. The same applies once your client is in custody; jail call transcription keeps those recordings searchable and citable too.

Salerno's warning is worth repeating to any court that will listen: if people learn officers use their behavior in the worst moments of their life against them, they'll stop calling 911 to help others.

Building the Challenge: The Scientific Record and the Motion Strategy

The scientific record against 911 call analysis is not a gray area, and the admissibility challenge writes itself once you assemble it. The original foundational study examined just 100 calls and was intended as exploratory research only, according to the Innocence Project's review of the method. At least eight replication studies have since failed to validate it, and researchers who tried to corroborate the creator's claims could not.

The institutional record compounds the problem. ProPublica's investigation documented more than 100 cases across 26 states where law enforcement employed the technique, and the University of Michigan Ford School STPP program published a policy brief framing it as lacking robust empirical support and posing an ongoing risk to impartial outcomes. Even the Colorado Bureau of Investigation conceded the method should be "nothing more than a potential investigative lead," as reported by Governing.

That framing matters because prosecutors rarely offer call analysis as expert testimony anymore. They route it through trained detectives and dispatchers as lay observation, so jurors hear guilt inferences without ever seeing the disputed methodology. Screen for the language in every report and proffered statement:

• Characterizations of the caller's emotional level ("flat," "calm," "not emotional enough")
• Claims the caller's phrasing or cadence indicated guilt, deception, or consciousness of guilt
• Detectives described as trained or certified in call analysis
• Opinions that the call was "inconsistent with" a genuine emergency or grief response

When you find it, confront the characterization against the actual audio. A word-timestamped transcript of the 911 call transcript analysis lets you show the jury what the caller actually said and when, which is the only effective answer to a flat-affect inference. The same discipline applies to jail call transcription, where officers frequently make similar behavioral claims.

Closed cases deserve a second look too. Fair and Just Prosecution has called on its members to review past convictions, and North Carolina's Office of Indigent Defense Services issued a statewide warning telling attorneys to watch for appearances of the technique. If a Harpster-trained analyst or a call-analysis-certified detective testified in one of your old files, that case belongs on your post-conviction list.

Confronting the Analysis with the Actual Audio: A Timestamped-Transcript Workflow

When the officer tells the jury your client "did not seem emotional enough," you need the exact seconds of the 911 call on the screen behind him. That flat-affect suspicion is a documented, research-validated bias: Dr. Jessica Salerno's studies with real 911 calls found civilians and police alike become more suspicious of low-emotion callers regardless of whether the caller was later convicted or exonerated. A timestamped transcript is how you put the call's actual words against the characterization.

Start with an accurate, word-timestamped, diarized transcript of the 911 audio. Not a rough summary. Not a typed-up paraphrase. Word-level timestamps and speaker labels, so every claim you make on cross is anchored to a playback position you can hit in the courtroom.

Then build the confrontation from the transcript itself:

• Mark the exact seconds the officer's report characterizes ("caller was calm," "no tears," "asked about the victim's condition before asking about help") and pull the transcript window around each one.
• Note what's actually said in those seconds: questions asked of dispatch, instructions given to a bystander, repetitions, silences. Cadence and content live in the transcript, not in the officer's summary.
• Flag any statement in the police report or witness interview that the call audio contradicts, and keep both quotes side by side.
• Draft cross questions tied to quoted sources only. Each question should point to a transcript moment, not to an inference.
• Play the seconds. Let the officer hear his own characterization against the call.

This is the workflow Evidence Copilot is built around. Upload the 911 audio with the rest of the discovery dump and you get a cited, diarized transcript with word-level timestamps; 911 call transcript analysis surfaces the transcript moments that matter, and review signals flag candidate inconsistencies between a statement and the call audio before you ask. Cross-exam drafts come out tied to quoted sources, saved as attorney-review notes. The same intake handles jail call transcription, so statements your client made in custody sit in the same review stream as the 911 call.

One caution worth naming: transcription tooling claims vary widely. Berkeley Law's resource library flags vendor-reported accuracy numbers as unverified and tells offices to evaluate tools independently before use. Treat every transcript as a working document you verify against the audio before you put it in front of a witness.

Every AI output in Evidence Copilot is marked AI-generated and pending human review, and every answer cites the exact second of audio it came from. That framing matters here more than anywhere: the cross works because you, the attorney, verified the transcript against the call and made the strategic call on what to confront. If you want to see the workflow on a real case file, book a 20-minute walkthrough or start on Solo.