What 911 Call Analysis Claims and Why It's Junk Science

You are likely facing a witness who claims to detect guilt from a 911 caller’s tone. This technique, known as 911 call analysis, lacks scientific validity but persists in courtrooms because it bypasses standard expert testimony rules.

The method originated in a 2006 master’s thesis by Detective Tracy Harpster, which analyzed only 100 exploratory calls. A co-author later admitted the findings were not ready for real-world application. Despite this weak foundation, police departments continue to use the technique to infer culpability from words and phrasing.

The core tool is the COPS Scale, or Considering Offender Probability in Statements. It categorizes caller behavior into six areas, including passive defenses and resistance to dispatchers. Prosecutors often cite specific "indicators" such as saying "Huh?" or failing to immediately plead for help.

• Original study: 100 calls, exploratory only.
• Replication: At least eight subsequent studies failed to validate the method.
• Adoption: Documented in at least 26 states and over 100 cases.
• Cost: Training programs cost up to $3,500 per officer.

The danger for your defense is structural. Because the method enters through lay testimony from trained detectives rather than expert testimony, it largely escapes Daubert or Frye challenges. Courts in Alabama and Nevada have rejected Harpster and trained detectives as experts, yet the practice continues. ProPublica found that some prosecutors knew the method was not recognized as scientific evidence but disguised it in trial anyway.

You must anticipate the counter-arguments. Harpster disputes the scientific critiques, claiming he has examined over 2,500 equivocal death cases. He argues the technique has exonerated innocent persons and is unaware of any wrongful conviction based on it alone. He points to case-specific evidence, such as financial motives or prior searches, as independent proof in cases like People v. Logan.

Treat these rebuttals as contested claims. The academic consensus, including research from Cornell University, shows that low emotion drives suspicion regardless of actual guilt. This bias shapes investigations and can lead to wrongful convictions, such as the case of Robert Roberson, who spent over 20 years on death row partly due to his flat affect.

To prepare effectively, you need the exact words and timing of the call. A timestamped transcript allows you to jump to specific moments and challenge the detective’s characterization. Tools like 911 call transcript analysis help you review the audio with precise citations. This workflow ensures you are arguing from the record, not the witness’s interpretation.

Consider how this fits into your broader discovery review. If the case involves jail calls, jail call transcription provides the same timestamped precision for those recordings. You need a grounded, searchable record to spot inconsistencies between the call and the police report.

The defense bar is mobilizing. NACDL’s president called the practice "dangerous and insidious." Post-conviction review efforts are underway, and at least one conviction in Michigan was overturned for improper admission. Know the COPS Scale indicators, challenge the lay testimony structure, and rely on the transcript, not the testimony.

How the Technique Slips Past Daubert as Lay Testimony

The most dangerous thing about 911 call analysis isn't its bogus science. It's the delivery mechanism. Prosecutors don't put Tracy Harpster on the stand and ask a court to qualify him as an expert, which would trigger a Daubert hearing and expose the method's total lack of validation. Instead, they call a Harpster-trained detective or dispatcher as an ordinary lay witness and let that witness narrate the call to the jury.

Christopher Lau, co-director of the Wisconsin Innocence Project, calls this the technique's defining feature: it "operates in the shadows," with the founders never proffered as experts, which lets it "largely escape Daubert or Frye evidentiary challenge." Harpster himself advises prosecutors trained in his method to guide juries through both the audio and the transcript of the call without formally qualifying as experts. The witness never offers an "opinion" in the FRE 702 sense. The jury just hears a police officer walk the call line by line, pausing on the phrases that sound suspicious, and draws the intended inference on its own.

The pattern is well documented. ProPublica's investigation found some prosecutors knew 911 call analysis would never survive admission as scientific evidence but introduced it anyway through lay testimony. The training behind these witnesses can run as little as eight hours and cost up to $3,500, often paid with taxpayer dollars. Eight hours. That's the entire foundation for testimony that has helped put people in prison for decades.

Here's what to watch for when the State's witness takes the stand:

• The witness "walks the jury through" the audio and transcript in narrative fashion rather than answering narrow factual questions about what was said.
• The testimony highlights COPS Scale indicators dressed up as common-sense observations: the caller said "Huh?" or "I don't know," used an awkward phrase, or failed to immediately plead for help for the victim.
• The witness's characterization of tone or emotion goes beyond anything a lay person could offer without the method, such as explaining that low emotion suggests guilt.
• The witness's training is mentioned glancingly, if at all, so the defense has no clean record on which to mount an expert-qualification challenge.

Each of these is an opportunity. If the witness is really applying the COPS Scale, that's specialized knowledge and the court should hold a Daubert hearing before a single indicator reaches the jury. Courts have already rejected Harpster himself as an expert in Alabama in 2009 and rejected a trained detective's testimony in Nevada. In Michigan, a 16-year-old's murder conviction was overturned because 911 call analysis should never have been admitted. The lay-testimony route is the workaround, and it fails when you name it out loud.

Preparation is what makes the objection land. You need the exact words and timestamps of the call in front of you so you can catch the moment the witness's "narration" drifts into method. That's where a transcript-first review workflow earns its keep; tools like Evidence Copilot's 911 call transcript analysis give you word-level timestamps and citations into the exact second of audio, so the witness's paraphrase gets checked against the record in real time. The same discipline applies to custodial phone evidence, which is why defense teams also lean on jail call transcription when the State's theory leans on a defendant's words.

Dr. Jessica Salerno's research adds the scientific backstop: when she played identical call scripts varying only in emotion, both civilians and police grew significantly more suspicious of low-emotion callers, and the pattern held with real calls regardless of whether the caller was later convicted or exonerated. The "indicator" the detective is walking the jury through is a suspicion bias, not a guilt signal. Your job at trial is to make the court see that the lay witness is carrying a discredited methodology through the side door, and to close that door before the jury hears a single "Huh?" characterized as evidence of anything.

The Cases That Show What's at Stake

Huwe Burton was 16 when a detective decided he "did not seem emotional enough," a judgment that helped convict him of his mother's murder. He spent more than 20 years in prison before his 2019 exoneration, according to the Innocence Project.

Burton's case isn't an outlier. Robert Roberson's flat affect after his daughter's death was read as guilt; he was later diagnosed with autism spectrum disorder, which explained the behavior. He has spent over 20 years on Texas death row and narrowly avoided execution three times.

Michigan shows what a successful challenge looks like. Riley Spitler's murder conviction was overturned outright because 911 call analysis should never have been admitted, as documented in the University of Cincinnati Law Review. In Illinois, Jessica Logan's conviction drew fresh scrutiny after ProPublica's reporting, and the state Supreme Court agreed to re-examine her case.

The common thread is emotion, not evidence. Cornell psychologist Jessica Salerno found that when civilians and police heard identical call scripts with varying emotion, both groups grew significantly more suspicious of low-emotion callers. For police, only highly emotional callers escaped that suspicion, and follow-up work with real calls showed the same pattern whether the caller was later convicted or exonerated.

That finding matters in cross-examination for three reasons:

• It reframes the detective's testimony as a reaction to demeanor, not a reading of guilt.
• It explains flat or unusual affect through trauma and neurodivergence rather than consciousness of guilt.
• It undercuts the premise that innocence follows a predictable emotional script.

Salerno's warning cuts deep for defenders: if the public learns police use behavior in someone's worst moment against them, people will stop calling 911 to help others. Your job is to make the record show exactly that, which starts with having the call's precise words and timestamps in hand. A grounded 911 call transcript analysis gives you the exact moments to quote back, and the same approach applies to recorded jail call transcription when the state's theory shifts to statements made after arrest.

Your 2026 Playbook: Get the Call, Get the Transcript, Challenge the Witness

Two facts should shape how you handle any 911 call in 2026: the original study behind call analysis examined just 100 calls, and at least eight later studies failed to validate it. The technique survives anyway, documented in at least 26 states and more than 100 cases, largely because it enters court through lay testimony rather than expert testimony.

That procedural move matters more than the underlying theory. Prosecutors introduce the method through Harpster-trained detectives and dispatchers who "walk the jury through" the call, which lets it slip past Daubert and Frye review. Your job is to see that pattern coming and build a record against it.

Step one: get the audio and a timestamped transcript into your own hands early. Request the 911 audio and any CAD or dispatch records at arraignment, not after the preliminary hearing. A timestamped transcript lets you pin a witness to exact words at exact moments, which is the only reliable way to counter a detective's characterization of tone or phrasing. For call audio and related custodial recordings, 911 call transcript analysis and jail call transcription both produce timestamped, searchable transcripts with speaker labels, so you can jump to the second a witness quotes and compare it against the recording. Every AI output stays marked for human review; you make the call.

Step two: prep against the six COPS categories. The proprietary COPS Scale (Considering Offender Probability in Statements) sorts indicators into six buckets: who the call is about, what it's about, how it's made, passive defenses, resistance to dispatcher, and additional guilty indicators. The specific "tells" prosecutors cite include:

• Saying "Huh?" or "I don't know"
• Unintelligible comments or awkward phrasing
• Failure to immediately plead for help for the victim
• Passive defenses and resistance to the dispatcher
• "Defendant mentality" and other catch-all indicators

Know these cold, because they define your cross-examination targets. They also define your Daubert attack: the indicator count grew from 16 in the original thesis to 38 by the 2017 book, and Harpster and Adams have never publicly released their samples, methodology, or underlying data. Courts have rejected Harpster himself as an expert (Alabama, 2009) and rejected trained detectives (Nevada), and Michigan overturned Riley Spitler's murder conviction because the analysis should not have been admitted.

Step three: pin the witness to words, not impressions. Low emotion drives suspicion regardless of actual guilt. Cornell psychologist Jessica Salerno found both civilians and police grew more suspicious of low-emotion callers even when the call scripts were identical, and the pattern held whether the caller was later convicted or exonerated. That's the frame for Huwe Burton, exonerated in 2019 after 20-plus years, and Robert Roberson, whose flat affect was later explained by an autism diagnosis.

When a detective testifies that your client "didn't sound right," hand him the transcript and ask him to point to the line. Evidence Copilot's workspace is built for exactly that: grounded citations into the exact moment, human review on every output, published pricing with no sales wall.