The Shadow Forensics Problem

The most dangerous junk science in a 2026 courtroom isn't the kind you can cross-examine. It's the kind that never takes the stand as science at all. 911 call analysis fits that description exactly, and standard admissibility motions often sail right past it.

The technique claims a caller's words and phrasing, including how they ask for help, can reveal whether they're the perpetrator. Christopher Lau, co-director of the Wisconsin Innocence Project, calls it "particularly pernicious because it operates in the shadows": the underlying methods are never disclosed and the technique's founders are never proffered as experts. Instead, a 911 operator or patrol officer who sat through a training testifies that the caller sounded guilty. The result, Lau writes, is that the technique "largely escapes Daubert or Frye evidentiary challenge."

That escape hatch is the whole game. Detective Tracy Harpster, who created the method in 2006, recommends prosecutors use it to "guide juries through both the audio and transcript of the calls" to create lasting impressions without ever seeking expert qualification. You can't move to exclude an expert who was never offered as one.

The scope is not small. ProPublica documented more than 100 cases across 26 states where law enforcement used Harpster's methods, and he has consulted on at least 2,500 homicide investigations. Meanwhile, the methodology rests on a sample of roughly 200 calls, and neither Harpster nor Dr. Susan Adams has publicly released the underlying data.

What that means for your motion practice:

• Object the moment an officer starts interpreting what a caller's words "indicate," and demand the basis for that opinion on the record.
• Preserve the actual call audio and transcript. You cannot rebut an interpretation you haven't verified against the source.
• Check the transcript against the audio word by word, because indicators like saying "I need help" or opening with "hi" have been treated as guilt markers.
• Build your record so the appellate issue is preserved, the way it was in the Michigan case where Riley Spitler's murder conviction was overturned because the analysis should not have been admitted.

Institutional cover is arriving. North Carolina's Office of Indigent Defense Services issued a statewide warning to defenders, and the Illinois Supreme Court agreed to revisit the Jessica Logan case after ProPublica's reporting. Fair and Just Prosecution has called on its members to review past convictions.

To be fair to the other side: Harpster denies any innocent person has been convicted using the analysis and says it has been used to exonerate people. The Colorado Bureau of Investigation's own position is that the technique should be "nothing more than a potential investigative lead," which concedes it should never decide a case.

The practical problem is evidentiary, not rhetorical. You need the call in a reviewable, citable form before you can argue about what it means. A 911 call transcript analysis workspace that timestamps every line and cites back to the exact second of audio lets you show the jury what the caller actually said, not what a training course told an officer to hear. The same discipline applies to jail call transcription, where custodial calls get scrutinized the same way. Evidence Copilot builds that record with diarized transcripts and citations into the source, so every claim about a call traces back to the call itself.

Dismantling the COPS Scale

The COPS Scale asks juries to believe that politeness is evidence of murder. That's not a caricature; it's what the indicators actually say.

The scale, COPS standing for "Considering Offender Probability in Statements," grew out of Detective Tracy Harpster's 2006 master's thesis. According to the University of Cincinnati Law Review, it started with 16 indicators, moved to 18 in a 2009 paper, and landed at 38 indicators in the 2017 book. The entire edifice rests on a sample of just 200 calls, and Harpster and co-author Dr. Susan Adams have never publicly released how they obtained those calls or made the underlying data available for independent review.

The indicators themselves are worth knowing cold, because they're your cross-examination roadmap. In a Michigan case, saying "I need help" was used to identify a caller as a suspect. Being "too polite," including opening the call with "hi," was cited as a sign the caller was hiding something. The scale sorts caller behavior into six categories:

• Who the call is about
• What the call is about
• How the call is made
• Passive defenses
• Resistance to the dispatcher
• Additional "guilty" indicators

Notice what's missing: any control group of innocent callers in genuine crisis. Nothing distinguishes a "guilty" phrasing from ordinary panic, shock, or a caller who simply speaks politely to strangers.

The empirical record is blunt. A University of Michigan policy brief finds that the COPS Scale© and similar techniques lack robust empirical support despite their routine use by law enforcement, and that they often reinforce investigative confirmation bias while diverting attention from more reliable evidence. The brief goes further: the application of 911 call analysis has contributed to wrongful convictions and poses an ongoing risk to impartial legal outcomes.

Even the Colorado Bureau of Investigation has said the technique should be nothing more than a potential investigative lead, a framing that concedes it should never be decisive evidence at trial.

The practical problem is that the scale's weakness rarely gets tested in court. Because the methods are proprietary and Harpster is seldom proffered as an expert, the indicators arrive through lay testimony from a 911 operator or officer who attended a training. Your counter starts with the primary source: pull the call audio, verify the transcript against it word by word, and build a record of what your client actually said versus what the analysis claims those words mean. That's the core of 911 call transcript analysis, and the same discipline applies to custodial recordings through jail call transcription, where prosecutors run the same word-and-phrasing inferences.

For fairness, note Harpster's position: he says he's unaware of any case in which an innocent person was convicted using the analysis and that it has been used to exonerate people. The appellate record, including Michigan's Riley Spitler reversal, tells a different story about what courts will tolerate going forward.

Building the Transcript Record

The call is your best evidence, but only if you actually have it. Because 911 call analysis operates as "shadow forensics," with methods never disclosed and founders never proffered as experts, the fight happens in the transcript, not just in a Daubert motion, according to Cornell Law Review scholarship. That means your record-building starts the day discovery lands.

Get the raw audio first, and demand it in discovery if it's missing. The technique has been used in more than 100 cases across 26 states, and prosecutors use it to "guide juries through both the audio and transcript of the calls" without expert qualification, per the University of Cincinnati Law Review. You need the same materials the State has, plus any dispatcher screen logs and CAD records.

Next, verify the transcript against the audio yourself. Don't trust the State's transcription, because the whole technique turns on tiny details of phrasing. Documented indicators include treating "I need help" as a guilt signal in a Michigan case, and reading a caller who opens with "hi" as "too polite" and possibly hiding something. If the transcript drops a filler word or renders a mumbled phrase differently, the State's narration of the call changes meaning. A word-level timestamped transcript lets you check each disputed word against the actual second of audio.

Then build your citation record:

• Timestamp every phrase the State characterizes, so you can quote the caller's exact words at 0:14 or 2:37 rather than a paraphrase.
• Note the dispatcher's questions alongside the caller's answers, since the COPS Scale treats "resistance to dispatcher" as a guilt category.
• Flag every place where the officer's report or testimony characterizes the call, versus what the audio actually contains.
• Record the elapsed time before the call, like the 17-minute delay litigated in the Jessica Logan case, so context survives into cross.

This is where a grounded workspace earns its keep. Evidence Copilot's 911 call transcript analysis gives you diarized, word-level timestamps with clickable citations into the exact second, so every quote in your cross outline points back to the audio. The same discipline applies to jail call transcription when the State is also mining recorded calls for phrasing.

Remember the foundation you're attacking: the COPS Scale rests on a sample of only 200 calls, and its developers have never released their data. When a trained officer takes the stand and narrates your client's tone, your timestamped record of what was actually said is the exhibit that lets the jury hear the difference.

Litigation Strategy and Precedent

The appellate landscape shifted in 2026, and defenders now have concrete wins to point to.

In Michigan, the Riley Spitler conviction was overturned because 911 call analysis should not have been admitted at trial, giving defenders a clean appellate precedent to cite in briefs and oral argument. That ruling matters because most challenges to this technique die in limine. The technique's underlying methods are never disclosed and its founders never take the stand, so it largely slips past standard Daubert or Frye scrutiny. Instead, 911 operators or officers testify about caller culpability after attending a Harpster training session.

This "shadow forensics" problem means a defense attorney who walks in expecting to win on a motion to suppress is already behind. The workaround is documented: prosecutors use the analysis to "guide juries through both the audio and transcript of the calls" to create "lasting impressions" without seeking expert qualification. Your objection strategy has to anticipate that entry point, not just the formal expert disclosure that rarely comes.

Building a record that survives appeal

Start with the call itself. Obtain the 911 audio and transcript, verify the transcript against the audio word by word, and document every moment where the analysis diverges from what the caller actually said. That grounded record is what makes a Spitler-style reversal possible.

Specific indicators worth pinning to the transcript:

• Saying "I need help" flagged a caller as a suspect in a Michigan case.
• Being "too polite," like opening with "hi," was cited as suggesting concealment.
• The COPS Scale rests on a sample of only 200 calls, with Harpster and Dr. Susan Adams never releasing how the research calls were obtained or making the data public.
• Indicators ballooned from 16 factors in the 2006 thesis to 38 in the 2017 book, with no peer-reviewed validation of the expanded set.

Institutional cover for the motion

North Carolina's Office of Indigent Defense Services issued a statewide warning to defense attorneys about the technique. Fair and Just Prosecution called on member prosecutors to review past cases through post-conviction processes. The Illinois Supreme Court agreed to revisit the Jessica Logan case after ProPublica's reporting, with the Exoneration Project and Center for Integrity in Forensic Sciences taking up representation. Each of these gives you authority to attach to a suppression motion or a post-conviction petition.

The Colorado Bureau of Investigation's own position helps too: the technique should be "nothing more than a potential investigative lead." When the agency that trained some of these officers concedes the ceiling, your argument writes itself.

The rebuttal you should expect

Harpster denies any innocent person has been convicted using his analysis and claims the method has exonerated people. Present that fairly in your briefing. Then note that the same critical consensus comes from the Cornell Law Review, the University of Michigan policy brief, the University of Cincinnati Law Review, and NACDL's framing of the technique as another example of forensic junk. One proponent's email correspondence does not outweigh the institutional record.

For defenders handling cases that also involve custodial calls, the same review discipline applies to jail call transcription, where grounded timestamps let you challenge any inference a witness tries to draw from a defendant's recorded statements. When you're ready to operationalize this workflow, Evidence Copilot's grounded workspace gives you timestamped diarized transcripts with citations into the exact second of audio, so every challenge you make in court ties back to a verifiable moment in the recording.