How 911 Call Analysis Sneaks Into Your Case Without an Expert

The prosecution doesn't need to call Tracy Harpster to get 911 call analysis into your trial. They call the detective who sat through his weekend seminar and let that officer walk the jury through the call as a "training and experience" witness. That's the end-run, and it's why the technique never faces Daubert or Frye scrutiny.

The Cornell Law Review calls this "shadow forensics": the methodology stays undisclosed and the founder never takes the stand, so courts never rule on whether the science holds up. The University of Cincinnati Law Review puts it bluntly in its title: "Unproven, Unreliable, Inadmissible," and notes Harpster has never been qualified as an expert in a criminal trial.

What actually lands in front of the jury is the COPS Scale (Considering Offender Probability in Statements), a proprietary rubric that scores callers across six categories: who the call is about, what it's about, how it's made, passive defenses, resistance to the dispatcher, and "additional guilty indicators." It grew from 16 indicators in Harpster's 2006 thesis to 38 in his 2017 book.

• Being "too polite": one mother's first word, "hi," was flagged as a guilt indicator
• Saying "Huh?" or "I don't know," or offering extraneous information
• Requesting police instead of an ambulance, or saying "blood" instead of "bleeding"
• Asking "Should I touch him?" or commenting on the victim's eyes
• Failing to make a desperate plea for help, or accepting the victim's death

ProPublica documented more than 100 cases in 26 states where law enforcement deployed these methods, and found prosecutors who knew the technique wouldn't survive expert qualification disguised it at trial anyway. An Orange County detective even emailed Harpster that the training gave the DA "suggestions on how to introduce the 911 calls to the jury during trial."

The damage is concrete. Huwe Burton served more than 20 years after an officer testified he "did not seem emotional enough" on the call; he was exonerated in 2019. Robert Roberson sat on Texas death row with a "flat affect" treated as suspicion until he was diagnosed with autism spectrum disorder. In Michigan, 16-year-old Riley Spitler's murder conviction was overturned because statement analysis never should have been admitted.

Screen for this in discovery. Ask whether any testifying officer or dispatcher attended Harpster training; North Carolina's Office of Indigent Defense Services issued a statewide warning telling defenders to do exactly that. A grounded 911 call transcript analysis with timestamped, diarized transcripts lets you put the officer's characterization next to the caller's actual words, second by second. Pair it with jail call transcription in the same workspace and you can document exactly what your client said versus what the analyst claims he meant.

The Science Doesn't Hold: Your Daubert/Frye Ammunition

The entire technique traces back to a 2006 master's thesis built on just 100 911 homicide calls, later expanded to 200. Harpster and co-author Dr. Susan Adams have never released their data or methodology publicly, which the University of Cincinnati Law Review calls contrary to standard scientific practice. The sample skews heavily Ohio, excludes intoxicated callers, covers only English-speaking Americans, and has unclear randomization and blinding, per McGill's science-review analysis.

That's your Daubert problem for the prosecution before the witness even takes the stand. Walk the factors:

• Testability and error rate: at least eight replication attempts have failed to validate the method, per the Innocence Project, with small-sample studies by Cromer (50 calls), Miller et al. (175 calls), and O'Donnell et al. (70 calls) producing mixed results at best.
• Peer review and general acceptance: courts have repeatedly rejected 911 call analysis as expert testimony, including a 2009 Alabama ruling excluding Harpster himself, and Harpster has never testified as an expert in a criminal trial.
• Internal contradiction: even FBI co-author John Jarvis told ProPublica he was "uncomfortable with 911 call analysis being used in real cases," saying "there's no definitive answer as to whether this is useful."
• Moving target: the indicator count grew from 16 (2006 thesis) to 18 (2009) to 38 (2017 book), which undercuts any claim of validated, stable standards.

The 2026 Cornell Law Review frames this as "shadow forensics": the method evades Daubert/Frye precisely because Harpster is never proffered, so seminar-trained detectives smuggle it in on "training and experience." Your motion should name that mechanism explicitly.

For cross-examination, arm yourself with Dr. Jessica Salerno's research at Cornell. Using identical 911 scripts, she found civilians and police became significantly more suspicious of low-emotion callers, and police required callers to sound highly emotional to reduce suspicion. Her warning: "Mistaken suspicion is rarely corrected by the investigation that follows. More often, it shapes the investigation itself."

Pair that with the disability and trauma counter-science. The technique has no answers for callers with autism, cognitive impairments, speech impediments, or non-native English. Robert Roberson spent over 20 years on Texas death row partly because his "flat affect" drew suspicion; he was later diagnosed with autism spectrum disorder. Huwe Burton served 20-plus years after an officer testified he "did not seem emotional enough," and was exonerated in 2019.

When you build the challenge file, you need the caller's actual words against the analyst's characterization, second by second. A grounded 911 call transcript analysis workflow with timestamped, diarized transcripts makes that comparison citable, and pairing it with jail call transcription in the same workspace lets you document phrasing across every recorded call in the case. The state is resting guilt on word choice; make them defend every word.

The 2026 Backlash and How to Use It

The institutional ground is shifting under 911 call analysis, and 2026 is the year to use that shift. ProPublica's investigation documented more than 100 cases across 26 states where law enforcement applied the technique, and the response has moved from academic criticism to actual consequences: bans under consideration, post-conviction reviews, and at least one conviction reversed outright.

The headline developments to track:

• The Illinois Supreme Court agreed to review Jessica Logan's conviction, which rested partly on a Harpster-trained detective's testimony about her 911 call. Logan is represented by the Exoneration Project and the Center for Integrity in Forensic Sciences.
• North Carolina's Office of Indigent Defense Services issued a statewide warning telling attorneys to watch for the technique in their cases.
• Fair and Just Prosecution, joined by the Innocence Project, called on member prosecutors to conduct post-conviction reviews of cases where 911 call analysis was used.
• In Michigan, Riley Spitler's murder conviction was overturned because 911 call analysis and other statement analysis should never have been admitted.

Each development does different work for you. If your client's case is pretrial, the Spitler reversal and the pending Logan review are your framing for a motion to exclude: courts are already willing to say this material doesn't belong in front of a jury. Move to bar any testimony that maps the caller's word choice, cadence, or emotional tone onto guilt, whether it comes in as expert opinion or as "training and experience" lay testimony.

If your client is already convicted, the calculus changed in 2026. Fair and Just Prosecution's review call and the Logan grant give you institutional cover that didn't exist two years ago. A post-conviction petition can now point to a sitting supreme court reviewing the exact issue, a national prosecutors' organization asking its own members to re-examine convictions, and NACDL's president calling the practice "dangerous and insidious." File while that momentum is live.

Expect the "investigative tool only" defense, because Harpster and his supporters deploy it in every forum. Susan Medina of the Colorado Bureau of Investigation has said her agency never advocated for 911 call analysis as a deciding factor in arrest or conviction. Your answer is the documented record: ProPublica found prosecutors who knew the technique wouldn't qualify as scientific evidence disguising it in trials anyway, and an Orange County detective emailed Harpster thanking him for "suggestions on how to introduce the 911 calls to the jury during trial." The tool-only framing collapses when the training materials themselves coach trial presentation.

Practically, that means discovery is where you win or lose this. Ask whether any testifying officer or dispatcher attended Harpster training, pull the seminar handouts, and get a verbatim, timestamped transcript of the call so you can hold the officer's characterizations against the caller's actual words. That comparison work, whether you do it manually or in a grounded workspace like Evidence Copilot that cites the exact second of audio, is what turns a vague objection into a record you can hand a judge.

A Working Workflow: Transcribing and Auditing the Calls

The whole fight over 911 call analysis comes down to one question: what did the caller actually say? If your only record is the officer's paraphrase at trial, you lose. If you have a timestamped transcript and can quote the exact words back, you have a cross-examination.

Start with the raw audio. Request the original 911 recording, not just the dispatcher's summary, and pull any jail calls in the discovery dump. Then produce verbatim transcripts with speaker labels and word-level timestamps for both. This matters because 911 call analysis rests entirely on word choice and phrasing, so a paraphrased transcript lets the analyst's interpretation substitute for the caller's actual words. A 911 call transcript analysis workflow with diarized, timestamped transcripts lets you compare the officer's characterizations against the caller's words second by second, and the same holds for jail call transcription when the state builds its narrative from later calls.

Next, run a word-by-word comparison. Take the officer's trial testimony line by line: "the caller never asked for help," "she accepted the death," "he was too calm." Then find the exact moment in your transcript that supports or contradicts each claim. Remember what the data shows: indicators like saying "hi" first, asking "Should I touch him?", or requesting police instead of an ambulance have been treated as guilt signals, and the underlying study started with only 100 calls and failed at least eight replication attempts. When the officer says the caller "didn't plead desperately," your transcript might show a plea the officer never mentioned.

Before any of that, screen for Harpster training in discovery. The technique reaches juries through seminar-trained detectives testifying on "training and experience," not through the creator himself, who has never taken the stand as an expert in a criminal trial. North Carolina's Office of Indigent Defense Services warned attorneys statewide to watch for exactly this. Ask:

• Did the testifying officer or dispatcher attend a Harpster seminar or any statement-analysis training?
• Who paid for it, and was it the roughly $3,500 eight-hour class documented in reporting?
• Does the officer's report or testimony use COPS Scale vocabulary like "passive defense" or "guilt indicators"?
• Did the state disclose any written analysis of the 911 call?

A grounded workspace keeps this manageable at scale. When every answer cites the exact second of audio it came from, you can build the comparison chart for cross-exam without re-listening to forty minutes of call audio to find one phrase. Every output stays marked AI-generated and pending your review, which is how it should be: the transcript is the evidence, the citations are the roadmap, and the judgment about what to ask on cross stays with you.