What 911 Call Analysis Actually Claims, and Why Courts Haven't Stopped It

The technique has a name, a proprietary scale, and a paper trail, and the paper trail starts with a study nobody should have relied on. In the mid-2000s, Ohio police lieutenant Tracy Harpster analyzed 100 recorded 911 calls, 50 from callers later deemed guilty and 50 from callers deemed not guilty, and published the results as exploratory research. A co-author on that work, FBI behavioral scientist John Jarvis, later said plainly that the work doesn't "rise to the level of success by the scientific community," yet the FBI distributed the study to roughly 200,000 law enforcement readers without labeling it exploratory, according to Forensic Magazine's investigation with ProPublica.

From those 100 calls grew a system: the COPS scale, short for Considering Offender Probability in Statements, which flags caller word choice and cadence as guilt indicators. The indicator count drifted over the years, from 16 in the 2006 thesis to 38 in the 2017 book, per the University of Cincinnati Law Review. Real cases turned on details like saying "hi" too politely or "I need help" sounding wrong.

The validation never arrived. Five independent studies, including two from the FBI's own Behavioral Analysis Unit plus researchers at Villanova and James Madison, failed to corroborate the method, and the FBI's 2020 study found indicators were so inconsistent that some pointed in the opposite direction of the original findings. The Innocence Project counts at least eight failed replication studies. The cost has been measured in years:

• Russ Faria served 3.5 years before acquittal at a second trial, where the judge cut short the 911 analysis testimony.
• Huwe Burton was convicted at 16 after an officer said he didn't seem emotional enough; he served over 20 years and was exonerated in 2019.
• Robert Roberson, diagnosed with autism after his flat affect was read as guilt, has spent more than 20 years on Texas death row.

Despite that record, ProPublica documented more than 100 cases across 26 states where the method shaped arrests, prosecutions, or convictions. Training runs up to $3,500 per eight-hour class, usually taxpayer-funded, and a survey of roughly 50 departments and training associations drew only 11 responses, which tells you how little anyone has tracked the spread.

So why do courts keep letting it in? Because it rarely arrives as expert testimony. Prosecutors route it through lay opinion under FRE 701, with detectives and dispatchers who attended a call analysis class testifying about caller culpability, and through prosecutor-guided walkthroughs of call audio. The Cornell Law Review calls this pattern "shadow forensics": the underlying methods are never disclosed, no one is proffered as an expert, and the technique escapes Daubert and Frye review entirely. One Missouri prosecutor put it in writing, describing how she got "creative" without calling it science.

For your cases, the practical move is to screen every 911 call at intake for the technique's fingerprints, then challenge any lay testimony that rests on specialized knowledge it can't claim. Reviewing call audio against what a testifying detective says the caller said takes timestamped, diarized transcripts you can cite back to the exact second, which is where 911 call transcript analysis and jail call transcription in a grounded review workspace earn their keep.

How to Spot the Technique in Your Discovery, Even When Nobody Calls It Science

Most defense attorneys never see 911 call analysis coming. A Virginia public defender told reporters she'd "never heard of any of that claptrap" in her jurisdiction, and dozens of defenders echoed the same reaction. That gap matters, because the technique doesn't announce itself in a discovery index. It shows up in the way a detective phrases his opinion of your client's "culpability," in the questions a prosecutor walks a jury through on the call audio, and in dispatcher testimony about who sounded "suspicious."

The reason it's so hard to spot is the workaround. Cornell Law Review calls this pattern "shadow forensics." The underlying methodology never gets disclosed, and its founders never get proffered as experts. Instead, a detective or dispatcher who attended a call-analysis training class testifies under FRE 701 about whether the caller sounded guilty. The technique escapes Daubert and Frye review entirely.

So the recognition work has to happen at intake, before anyone walks into a courtroom. Screen every 911 recording and transcript for these fingerprints:

• Detectives or dispatchers testifying about caller "culpability" after attending call-analysis training
• Prosecutor-guided walkthroughs of the call audio, where the State cherry-picks moments for the jury
• Dispatcher testimony flagging a caller as "suspicious" before any investigation began
• COPS-scale indicators in the record: a greeting flagged as "too polite," a plea like "I need help" reframed as incriminating

The UC Law Review documents these indicators in the proprietary "COPS" scale (Considering Offender Probability in Statements), organized into six categories with as many as 38 red flags. A "hi" on the call can be marked suspicious. So can "I need help."

What makes this dangerous is the confirmation-bias cascade it triggers. Once a detective decides a caller sounds guilty, that suspicion shapes every investigative step that follows. The Innocence Project puts it plainly: mistaken suspicion is rarely corrected by the investigation that follows. More often, it shapes the investigation itself.

The emotional-affect research from Dr. Jessica Salerno at Cornell sharpens the problem. In controlled studies using identical 911 scripts with varying levels of caller emotion, both civilians and police became significantly more suspicious of low-emotion callers. For police, even moderate emotion wasn't enough. Callers had to sound highly emotional before suspicion dropped. Follow-up work using real 911 calls with known outcomes found the same low-emotion-equals-guilt association, regardless of whether the caller was later convicted or exonerated.

That research has a neurodivergence dimension you can put in front of a jury. Robert Roberson's flat affect on a 911 call about his daughter's death was read as guilt. He was later diagnosed with autism spectrum disorder and has spent more than 20 years on Texas death row.

Spotting the technique means treating every 911 transcript the way you'd treat a lab report: read it against the detective's testimony, line by line, and flag every opinion that traces back to a training class rather than evidence. A timestamped, diarized transcript with citations back to the exact audio moment makes that comparison possible. That's the workflow a service like 911 call transcript analysis is built for, and the same approach applies when you're screening jail call transcription for the same indicators.

Building the Record: Grounded Transcript Review and the FRE 701/702 Challenge

When a detective takes the stand and tells the jury the caller "didn't act like a victim," your objection lives or dies on the record. If you can't point to the exact second the call shows something different, the state's narrative wins. That's why building the record comes before the legal argument: you can't impeach testimony you can't quote.

The legal move is well documented. Because 911 call analysis can't survive Daubert or Frye scrutiny, prosecutors introduce it indirectly, embedding the disputed methodology in lay opinion testimony from trained detectives or dispatchers. Cornell Law Review calls this pattern "shadow forensics": the underlying methods are never disclosed, the founders are never proffered as experts, and the technique "largely escapes Daubert or Frye evidentiary challenge." Your FRE 701 argument is that lay testimony cannot rest on scientific, technical, or specialized knowledge, and the caller-culpability inferences are exactly that.

The reliability record is devastating, and it belongs in your motion:

• Five independent studies failed to corroborate the method, including two from the FBI's own Behavioral Analysis Unit, which found indicators so inconsistent that some went "in the opposite direction of what was previously found."
• The Innocence Project counts at least eight failed replication or validation studies of the technique.
• Co-author John Jarvis, an FBI behavioral scientist, admitted the work doesn't "rise to the level of success by the scientific community" and that there's "no definitive answer as to whether this is useful."
• Harpster has never released his underlying data or methodology, despite claiming to have consulted on 2,500+ cases, and has never testified as an expert himself.

That last point matters at the hearing. A method whose creator won't produce his data has no error rate, no validation studies you can audit, and no general acceptance to point to. As the UC Law Review details, courts have repeatedly rejected the technique when it was actually presented for what it is.

Then comes the evidentiary groundwork. The detective will paraphrase. "The caller stated she needed help." "He was unusually polite." Your job is to have a timestamped, diarized transcript of the actual call so you can cite back to the exact moment and show what the caller really said. This is where 911 call transcript analysis and jail call transcription do the load-bearing work: surfacing the gap between what a testifying detective claims the caller said and what the recording actually shows. Tools like Evidence Copilot produce word-level timestamps with speaker diarization, so every cross-exam question ties to a quoted, citable moment rather than your memory of the audio.

Screen every case with 911 audio at intake. Build the transcript record first, then file the 701/702 challenge with the failed studies attached.

Using the 2026 Momentum: Bans, Conviction Reviews, and What to Do Next

The ProPublica investigation didn't just document a junk science problem. It set off an institutional response that defense attorneys can use right now, in pretrial motions and post-conviction petitions alike.

The momentum is real and it's coming from every direction. 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 convictions, with president Miriam Krinsky presenting the reporting at a prosecutors' summit in Austin. The Illinois Supreme Court agreed to review Jessica Logan's case after the reporting broke. And NACDL president Nellie King called for adding the technique "to the junk pile of fake scientific theories" contributing to wrongful convictions, noting that "justice agencies... are buying what he is selling" (Governing).

That shift matters because, as the research shows, defense awareness has been the gap. A Virginia public defender told reporters he'd "never heard of any of that claptrap in my jurisdiction," and dozens of defense attorneys reacted the same way (Forensic Magazine). Dr. Jessica Salerno's advice to the bar is blunt: learn to identify and challenge attempts to introduce the technique through lay testimony. Here's what that looks like in practice:

• Screen every case with 911 audio at intake. Look for the COPS-scale fingerprints, like treating "hi" as too polite or "I need help" as suspicious, both documented in real cases.
• Watch for detectives or dispatchers testifying about a caller's "culpability" after attending call analysis training. That's the shadow-forensics workaround Cornell Law Review describes: the method escapes Daubert or Frye scrutiny because it's never disclosed as science.
• Build the impeachment record from the published studies: five failed corroboration studies including two from the FBI's own Behavioral Analysis Unit, at least eight failed replications overall, and Harpster's own co-author conceding the work doesn't rise to the level of scientific success.
• In post-conviction work, cite the statewide warnings, the prosecutorial review calls, and the Logan review as evidence that courts and institutional actors now treat the technique as unreliable.

The practical bottleneck is the audio itself. Comparing what a testifying detective claims the caller said against what the recording actually shows means hours of listening, timestamped transcripts, and a clean citation trail. That's the workflow Evidence Copilot is built for: 911 call transcript analysis and jail call transcription with diarized, word-level timestamps, every AI answer cited to the exact moment and marked for human review. Published pricing (Solo $39/user/mo, Firm $99/user/mo, PD Office $750/attorney/yr) means you can budget it into a case without a sales call.

The window is open. Prosecutors, judges, and the defense bar are all saying the same thing about this technique. The attorneys who move first, with the transcript work to back it up, will be the ones who get it excluded.