How 'Shadow Forensics' Slips 911 Call Analysis Past Daubert

Prosecutors have a workaround for 911 call analysis, and it does not involve defending the science. They put a dispatcher or detective on the stand as a lay witness, and the technique never faces a Daubert or Frye hearing at all.

Christopher Lau, co-director of the Wisconsin Innocence Project, calls this "shadow forensics": the underlying methods are never disclosed, and the founders of the technique are never proffered as experts. Writing in the Cornell Law Review, Lau argues that 911 call analysis is "particularly pernicious because it operates in the shadows," largely escaping the evidentiary challenge that would gut it. The officer testifies about caller culpability after an eight-hour training, and the jury hears an opinion dressed as observation.

The scale is not small. ProPublica documented more than 100 cases across 26 states where law enforcement used these methods, with the FBI helping the technique spread. Departments pay up to $3,500 for courses that promise to teach officers to ascertain guilt from a 911 call in a single day.

The methods themselves stay hidden, and what leaks out is troubling:

• The COPS scale and its guilt indicators were never released for independent review. The creator, Tracy Harpster, has not disclosed how calls were obtained or how the research was conducted.
• The original study examined only 100 calls and was intended as exploratory research, per the Innocence Project.
• At least eight subsequent studies tried to replicate the findings. None held up.
• The claimed indicators of guilt grew from 16 in a 2006 thesis to 18 in a 2009 paper to 38 in a 2017 book, with no validation study behind the additions.

Then there is the question the officer usually cannot answer: what is the science behind your opinion? The University of Cincinnati Law Review notes that Harpster has never testified as an expert in a criminal trial, and the method fails all five Daubert factors: testability, peer review, known error rate, controlling standards, and general acceptance. A lay witness who cannot explain reliability is testifying to a conclusion the court never vetted.

For your motion practice, the practical move is to force the issue. If the officer's "opinion" rests on specialized training and a proprietary scale, it is expert testimony, and the state must lay the Daubert foundation it has been avoiding. Pair that challenge with a clean, timestamped transcript so you can show the jury exactly what the caller said, and when. Defense teams running 911 call transcript analysis alongside bodycam and report review tend to find the delays and word choices the state characterizes as guilt are just as consistent with panic, shock, or a caller who never expected to be a suspect.

The Junk Science Record: Building Your Admissibility Attack

The empirical case against 911 call analysis is more one-sided than almost any forensic technique you'll challenge this year. The original study behind the method examined only 100 calls and was, by its own design, exploratory research never intended to validate courtroom conclusions, according to the Innocence Project.

It gets worse from there. At least eight subsequent studies have tried to replicate or test the method, and each one found the claimed guilt indicators don't hold up. As Miami defense attorney Julia Kefalinos puts it, no one has been able to replicate Harpster's findings, and it's unclear whether the analysts even knew case outcomes before listening to the calls.

The method itself keeps shifting under your feet. The claimed guilt indicators grew from 16 in the 2006 thesis to 18 in a 2009 paper to 38 by the 2017 book, per the University of Cincinnati Law Review. A technique that reliably measures something doesn't quintuple its indicator list over a decade. Meanwhile, creator Tracy Harpster has never testified as an expert in a criminal trial, has consulted on at least 2,500 homicide investigations, and has never released how his call samples were obtained or how the research was conducted.

Follow the money too. Departments pay up to $3,500 for an eight-hour course promising to teach officers to ascertain guilt from 911 calls. Eight hours. Your Daubert motion practically writes itself when you map the method onto the factors:

• Testability: the underlying data and methods have never been disclosed, so the claims can't be independently tested.
• Peer review and error rate: no published validation studies, and eight replication attempts failed.
• Standards and controls: the original studies lacked controls, and the COPS scale is proprietary.
• General acceptance: the technique fails Frye, and even law enforcement hedges. The Colorado Bureau of Investigation says it should never be a deciding factor in an arrest or conviction.

Expect the prosecution to route around all of this with lay testimony from a detective or dispatcher who took the training. Challenge that head-on: an opinion built on specialized training and a proprietary scale is expert testimony wearing a lay witness costume.

Then bring in the emotion research. Dr. Jessica Salerno's work shows officers become significantly more suspicious of low-emotion callers, and moderate emotion often isn't enough to change their perception. What investigators are reading as guilt, Salerno notes, is often just the absence of an expected display of emotion. Shock, trauma, and neurodivergence all produce flat affect. That's your cross-examination.

Finally, use the institutional record. The NACDL has called the technique forensic junk, prosecutors and judges nationwide have called for bans and post-conviction review, and ProPublica documented its use in more than 100 cases across 26 states. Cite all of it in your motion to exclude. When you're pulling the actual call audio, bodycam, and reports together for that filing, a grounded review workspace like Evidence Copilot keeps every quote tied to a timestamped citation, which is exactly what a Daubert hearing demands.

Challenging the Lay Testimony Loophole and Demanding Disclosure

The most dangerous thing about 911 call analysis isn't the science. It's the paperwork, or rather, the absence of it. Prosecutors have learned they can get caller-culpability opinions in front of a jury without ever triggering a Daubert hearing, and the loophole runs through one word: "lay."

Under the rules of evidence, a lay witness may opine about what they personally perceived. They may not opine based on specialized knowledge. Yet detectives and dispatchers routinely testify that a caller "didn't display the emotions of an innocent person" after attending a training course, sometimes one that costs departments up to $3,500 for a single eight-hour session promising to teach officers to infer guilt from 911 calls. As Christopher Lau of the Wisconsin Innocence Project has written, this framing lets the technique largely escape Daubert or Frye challenge because the "method" never gets named as a method. Scholars have dubbed the tactic "shadow forensics": the underlying methods are never disclosed, and the technique's creator, Tracy Harpster, has never testified as an expert in a criminal trial.

Your motion practice should attack that framing directly. The argument writes itself once you look at what the witness actually relied on:

• A proprietary scale (the COPS scale) the witness did not develop and cannot explain the origins of
• A list of "guilt indicators" that has ballooned from 16 in a 2006 thesis to 18 in 2009 to 38 in a 2017 book, with no published validation at any stage
• Training materials from a paid course, not personal perception
• Conclusions about deception and culpability, which are classic expert domains

That's specialized knowledge dressed up as observation. File a motion arguing the testimony is expert testimony masquerading as lay observation, and follow it with motions to compel: the specific indicators the officer applied, the training materials the officer received, and the underlying data. Demand the dataset itself. The original study examined only 100 calls and was intended as exploratory research, and at least eight subsequent studies have failed to replicate its indicators. Harpster has never released how calls were obtained or how the research was conducted. If the prosecution's witness can't produce the foundation, the "opinion" collapses into an officer telling the jury the defendant sounded guilty.

Pair the disclosure fight with your cross. Dr. Jessica Salerno's research shows officers become significantly more suspicious of low-emotion callers, and even moderate emotion often isn't enough to change their perception; callers must sound "highly emotional" to clear the bar. That flat affect the state calls suspicious? It's a documented response to shock and trauma, and it's common in neurodivergent callers, including people on the autism spectrum. Retain an expert on trauma and emotional expression and dismantle the premise that there's one "correct" way to sound while reporting a crisis. As Salerno puts it, what jurors and investigators are often detecting is simply "the absence of an expected display of emotion."

Practically, this fight lives and dies on the transcript. You need an accurate, timestamped, diarized transcript of the call to test every claimed indicator against what was actually said and when. When we review 911 call transcript analysis at Evidence Copilot, every quote ties back to the exact second of audio, so an officer's characterization of "hesitation" or "flat tone" can be checked line by line against the record. The same workflow applies when jail call transcription feeds a companion narrative; consistency between calls, bodycam, and reports is where these subjective guilt narratives fall apart.

Working the Call Itself: Transcript Review and Cross-Prep Workflow

The call audio is the one piece of evidence both sides quote, and the transcript you build from it decides who gets to be precise. Pull the 911 audio into a case, let diarization and word-level timestamps run, then rename speakers so every "Operator" and "Caller" tag stays consistent across segments.

From there, the review is mechanical. Read the transcript against the police report and the bodycam, second by second, and flag anything that doesn't line up. The documented cases show why this matters: in the Jessica Logan prosecution, the state's theory rested on a 17-minute delay before the call, a delay defense experts contested and calls for a ban on 911 call analysis now cite alongside more than 100 cases in 26 states. A timestamped transcript turns "about 17 minutes" into an exact span you can quote.

• Confirm the call's start and end times against CAD logs and the report's stated dispatch time.
• Check the caller's words against bodycam audio from the same window for contradictions.
• Note every gap, pause, or interruption with its timestamp, not a paraphrase.
• Flag any officer or detective summary that describes the call's content differently than the audio does.

Then build the cross. If a detective or operator offers an opinion about the caller's demeanor, your questions come straight from the record: what training did you receive, what indicators did you apply, and where in the audio does that indicator appear? The University of Cincinnati Law Review notes the method expanded from 16 claimed indicators in 2006 to 38 by 2017, which gives you plenty to pin down.

This is where 911 call transcript analysis earns its place in the workflow: every quoted moment stays cited to the exact second, and AI-generated output is flagged for attorney review rather than treated as a conclusion. Pair it with jail call transcription when the same client's recorded statements need the same treatment. You can book a 20-minute walkthrough to see the citation trail on your own case file.