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

If you defend criminal cases for a living, you're going to meet 911 call analysis in 2026, whether it arrives labeled or not. The technique claims a caller's word choice, emotion level, and call timing can reveal whether that caller is the perpetrator, and it has spread to more than 100 cases across 26 states.

The method's chief architect is Tracy Harpster, a retired Ohio detective who sells an eight-hour training program to police departments for up to $3,500 per session, taxpayer money included. His theory rests on roughly 200 total calls from two small studies, and it claims 20 variables in a 911 call can flag a guilty caller. Among the supposed indicators: whether a caller comments on the victim's eyes, asks whether to touch the body, says "blood" instead of "bleeding," or immediately acknowledges death.

The scientific record is thin, and defenders should know it cold. At least eight studies have tried to replicate or test the method, and the claimed guilt indicators do not hold up. The foundational study examined only 100 calls, and a co-author later admitted the findings were "not ready for real-world application."

Why it lands in your discovery anyway: prosecutors reportedly sidestep scientific-evidence standards by introducing the technique through lay testimony from trained detectives or dispatchers rather than expert testimony. Jurors then hear conclusions about "guilty" behavior without learning the methodology is disputed. Dr. Jessica Salerno of Cornell, whose research shows suspicion tracks absence of expected emotion rather than deception, has explicitly recommended that defense attorneys learn to identify and challenge these lay-testimony attempts.

That challenge starts with the audio itself. Before you can cross a detective on what a call supposedly reveals, you need a timestamped transcript you can search, quote, and line up against the police report and bodycam. For teams working through 911 audio at case scale, 911 call transcript analysis and jail call transcription are the two Evidence Copilot services that map most directly onto this problem: diarized transcripts with word-level timestamps, every AI answer cited to the exact moment, and human review marked on every output.

• Harpster's rebuttal deserves a fair read: he says he is "unaware of any case in which an innocent person was convicted using this analysis," and ProPublica has responded that it never reported the technique caused such a conviction.
• Supporters in law enforcement frame it as an investigative lead only. The Colorado Bureau of Investigation's chief of staff said the department "would never advocate for 911 call analysis to be a deciding factor of an arrest or conviction."
• Case law is still developing. Jessica Logan's conviction was upheld on appeal before the Illinois Supreme Court took up her case after ProPublica's reporting.
• NACDL published "911 Call Analysis: Another Example Forensic Junk..." in May 2026, but the article is members-only, so treat it as a pointer for members rather than a citable source.

The Scientific Record You Can Cite in a Motion or Cross

If you plan to challenge 911 call analysis in a motion or a cross, you need the record at your fingertips. The technique's foundational study examined only 100 calls and was intended as exploratory research only, and one co-author later told ProPublica the findings were "not ready for real-world application." Since then, at least eight scientific studies have attempted to replicate or test the method, and its supposed indicators of guilt keep failing to hold up under scrutiny.

The claim set itself is worth knowing cold: Harpster's theory holds that 20 variables in a 911 call can indicate whether the caller is the perpetrator, and it rests on two small studies of roughly 100 calls each. Training in the method sells to law enforcement for up to $3,500 per program, often funded by taxpayer money, promising that eight hours of instruction can teach dispatchers, detectives, or prosecutors to determine guilt from a call.

The strongest counter-evidence comes from Dr. Jessica Salerno at Cornell. Her studies, using identical 911 scripts with varying emotion levels, found that both civilians and police grew significantly more suspicious when callers expressed low emotion. For officers, even moderate emotion was insufficient; callers had to sound highly emotional before suspicion decreased. The low-emotion/guilt association held in follow-up work using real 911 calls with known outcomes, regardless of whether the caller was later convicted or later exonerated. That means the cue officers read as guilt tracks emotion, not deception.

Two cases show what that costs:

• Huwe Burton found his mother murdered in their Bronx apartment at 16, called 911, and was charged days later. At trial, an officer testified Burton "did not seem emotional enough." He was exonerated in 2019, after serving more than 20 years.
• Robert Roberson's flat affect after bringing his critically ill daughter to the hospital was read as evidence of guilt. He was later diagnosed with autism spectrum disorder, which authorities had not accounted for. He has spent more than two decades on Texas' death row.

Harpster disputes the framing. In a letter to ProPublica, he wrote, "I am unaware of any case in which an innocent person was convicted using this analysis. If fact, the analysis has been used to exonerate innocent persons," and ProPublica responded that it never reported the technique caused the conviction of innocent persons. Supporters in law enforcement, including the Colorado Bureau of Investigation's chief of staff, describe it as a valuable investigative tool but not decisive evidence on which to base a conviction. Critics, including NACDL president Nellie King, call the business "dangerous and insidious." NACDL members can find the May 2026 article "911 Call Analysis: Another Example Forensic Junk..." in the members-only library, though its content is not publicly verifiable.

None of this helps if you never actually hear the call. Public defenders report an average of just 7 minutes per case, and AI-assisted transcription is reported to cut review time by roughly 50%. A grounded workspace that produces timestamped 911 call transcript analysis and jail call transcription, with every answer citing the exact source moment, is how you get from a motion outline to a cross that quotes the call line by line.

The Lay-Testimony Workaround: How It Slips Past Daubert

The most dangerous version of 911 call analysis never appears on a witness list labeled "expert." Prosecutors knew the technique wouldn't survive scrutiny as scientific evidence, yet reporting by ProPublica documented cases where it was disguised in trial anyway. The mechanism is simple: a detective or dispatcher who took the eight-hour Harpster training takes the stand as a lay witness and testifies the caller "didn't sound like a grieving husband." Jurors hear a guilt conclusion. Nobody hears that at least eight studies have failed to replicate the method's indicators.

Dr. Jessica Salerno's research explains why this testimony is so persuasive and so hollow. Suspicion is driven by the absence of expected emotion, not by actual deception cues, and the low-emotion/guilt association held even for callers later exonerated. Huwe Burton spent more than 20 years in prison after an officer testified he "did not seem emotional enough." That's the whole evidentiary basis, dressed up as investigative experience.

Your spotting checklist, stage by stage:

• Witness lists: any detective or dispatcher listed as a fact witness in a homicide where the 911 call exists, especially if their report quotes the call's tone, cadence, or word choice rather than its content.
• Police reports and investigative narratives: phrases like "inconsistent with," " indicators of," or "the caller's affect" are training-program vocabulary, not natural report writing.
• Preliminary hearing transcripts: watch for lay witnesses who "noticed things" about the call. Object early so the record shows the challenge.
• Discovery requests: demand training records, course materials, and any Harpster correspondence. The Orange County detective's email praising consultation on "indicators of guilt" came out in reporting; similar documents exist in your case.

When you spot it, move for a motion in limine to preclude opinion testimony on the caller's demeanor, and argue the witness is offering a scientific conclusion without Daubert scrutiny. Cross on the training's origin: two exploratory studies of roughly 100 calls each, with a co-author conceding the findings were "not ready for real-world application."

You can't cross what you haven't heard, and you can't challenge a call you haven't reviewed word by word. Getting a timestamped transcript of the 911 audio, and the jail call transcription of your client's later statements, lets you compare what the caller actually said against the detective's characterization. Evidence Copilot's 911 call transcript analysis exists for exactly this cross-check, with every answer cited to the second of audio. The detective testifies to an impression. You show the jury the words.

Building the Review Workflow: Transcribing and Cross-Checking the Call

You cannot challenge a 911 call you have never actually heard, and most defenders never get the chance. The average amount of time some public defenders can dedicate to any one case is seven minutes. That is the workload reality sitting underneath every 911 call analysis challenge you plan to file.

The fix is a repeatable review workflow, not more hours. Start by getting the audio into a timestamped, speaker-labeled transcript so you can search it instead of listening to it in real time. AI-assisted transcription is reported to cut evidence review by roughly 50 percent, and tools built for defense work now transcribe 911 calls alongside bodycam and jail audio, per UC Berkeley's public defense AI tool library.

A workable sequence looks like this:

• Pull the 911 audio and confirm you have the complete call, not a clipped excerpt.
• Generate a diarized transcript with word-level timestamps, then rename the callers and dispatchers.
• Read the police report cold, then read the transcript against it line by line.
• Pull bodycam from the same window and sync it to the call.
• Flag every place the report, the call, and the footage disagree, and quote each side verbatim.

That cross-check is where cases move. A Harris County public defender described a case dismissed outright "because of what was on the body cam," after the offense report and the footage failed to line up, in the same Stand Together feature on defender workloads.

The same discipline applies to custody recordings. If the government plans to use a jail call to show consciousness of guilt, you need that audio transcribed and timestamped too, which is why teams route both through 911 call transcript analysis and jail call transcription rather than reviewing each recording by ear.

Evidence Copilot exists for exactly this stage: upload the discovery dump, get timestamped diarized transcripts and OCR'd reports, and let the evidence agent answer questions with citations that seek the exact second of audio or page of the PDF. Every AI output is marked pending human review, and originals stay untouched.

Your 2026 Action Checklist

You cannot challenge a 911 call analysis you have not thoroughly reviewed. The defense workflow for 2026 requires moving from passive receipt of discovery to active, systematic scrutiny of every audio file in the case.

Start by identifying potential bias in the personnel involved. If a detective or dispatcher claims expertise in 911 call analysis, verify their training background. Investigative reporting documents that more than 100 cases across 26 states have utilized methods developed by Tracy Harpster. Flag any officer who attended these specific training programs during your discovery review. This context is critical for assessing the credibility of their testimony.

Next, demand the methodology behind any conclusions presented. Prosecutors often introduce these claims through lay testimony rather than expert testimony to bypass scientific standards. Research from Cornell University indicates that suspicion is often driven by the absence of expected emotion rather than actual deception cues. Prepare your Daubert objections now. You must also prepare to challenge lay testimony that presents unvalidated behavioral indicators as fact.

Finally, run a transcript-versus-report cross-check on every 911 and jail call in the case. Do not rely on the police report alone. As one Harris County public defender noted, the offense report and the truth may not line up, leading to case dismissals only when the bodycam or audio evidence is reviewed. 911 call transcript analysis allows you to compare the caller’s exact words against the narrative in the police report. Similarly, use jail call transcription to verify statements made during booking against initial reports.

This process is time-intensive, which is why many defense teams struggle to keep up. Some public defenders report an average of just 7 minutes per case for review. To manage this workload, consider using a workspace that automates transcription and flags inconsistencies. Evidence Copilot provides a grounded evidence-review environment where you can upload audio files and receive timestamped transcripts with speaker diarization. The system automatically compares new evidence against existing files to surface potential inconsistencies, such as discrepancies between a call recording and a police report.

For a deeper dive into the legal arguments against this technique, NACDL members can access a May 2026 article titled "911 Call Analysis: Another Example Forensic Junk..." on their website. This resource offers additional context for building your suppression motions.

• Flag any detective or dispatcher who has received Harpster training in your discovery notes.
• Prepare Daubert challenges for expert testimony and lay-testimony objections for indirect claims.
• Cross-check every 911 and jail call transcript against the corresponding police report.
• Document any discrepancies between the caller's statements and the officer's narrative.