What 911 Call Analysis Is and Why It's Still Showing Up in Your Cases

You likely have a 911 call in discovery that the State intends to use against you. The problem is not the audio file itself. It is the "analysis" they will build on top of it.

911 call analysis infers guilt from word choice, phrasing, and cadence. It claims that specific verbal patterns prove a caller is the perpetrator. This technique has spread rapidly despite a total lack of scientific validation. As of 2024, it has influenced over 2,500 homicide investigations across at least 26 states.

The legal landscape is shifting fast. Law reviews and investigative journalists are now reclassifying this method as "junk forensics." The University of Cincinnati Law Review published an article titled outright "Unproven, Unreliable, Inadmissible." A Cornell Law Review piece coined the term "shadow forensics" to describe how the technique escapes Daubert scrutiny. The founders are never proffered as experts. The underlying methods are never disclosed. Instead, a police officer who attended a one-day training testifies about your client’s "guilt indicators."

This structure is designed to bypass evidentiary hurdles. Prosecutors often use a "lay testimony" workaround. They "guide the jury" through the audio, mixing the 38 indicators from the COPS Scale into argument rather than expert opinion. You need to understand the mechanics of this process to dismantle it.

The core data behind the technique is thin. It stems from a 2006 master’s thesis with a sample of only 100 calls. The current scale expands this to 200 calls but still lacks peer review. There is no published error rate. There is no independent replication. The Ford School of Public Policy notes that these methods often reinforce investigative confirmation bias.

You cannot fight a narrative you cannot see. If you rely on the State’s transcript, you are fighting their version of the facts. You need a word-level, speaker-separated record. This allows you to impeach the "analysis" against what the caller actually said. A timestamped, diarized transcript grounds every rebuttal in the specific second of audio.

Consider the strategic implications for your workflow:

• Demand the training materials. If the State’s witness attended a Harpster training, the curriculum and slide decks are discoverable. You need to know exactly which "indicators" they were taught to look for.
• Pre-empt the lay testimony trap. File a motion in limine that excludes any "walkthrough" of the call. If the prosecutor argues that a pause implies guilt, you must have a clean transcript to show the pause was due to breathing or a technical glitch.
• Preserve the post-conviction issue. Reversals are already happening. A Michigan conviction was overturned due to improper admission. The Illinois Supreme Court is reviewing the Jessica Logan case. Your objection at trial must be specific and on the record.

This technique is not going away on its own. Departments continue to pay for trainings. Prosecutors continue to use the results. You must treat every 911 call as a potential forensic battleground.

At Evidence Copilot, we help defense teams manage this complexity. Our workspace processes 911 audio into 911 call transcript analysis with word-level timestamps and speaker diarization. This ensures you can cite the exact moment a "guilt indicator" appears, or prove it does not exist. The same discipline applies to jail call transcription when the State tries to use post-arrest statements as consciousness of guilt. You need the raw data to control the narrative.

The Shadow Forensics Loophole: How the Technique Skips Daubert Entirely

The technique does not survive a Daubert challenge. It avoids one entirely.

That's the core finding from the Cornell Law Review's "shadow forensics" article, and it should change how every defender approaches a 911 call analysis case. The method's founders are never proffered as experts. The underlying methodology is never disclosed. Instead, 911 operators or police officers who sat through a training session testify about the caller's culpability.

The effect, according to Cornell, is that the technique "largely escapes Daubert or Frye evidentiary challenge." The gatekeeping hearing never happens because no expert takes the stand.

This is a disclosure and confrontation problem dressed up as a science problem. And prosecutors know it.

When expert testimony is unavailable, the University of Cincinnati Law Review documents that prosecutors use two documented workarounds. First, they introduce the analysis through lay testimony from officers who attended Harpster's trainings. Second, they "guide juries through both the audio and transcript of the calls" to "create lasting impressions of this critical evidence," a recommendation that comes straight from Harpster's own training materials.

Neither path triggers a reliability hearing. Neither requires the State to produce the underlying data. Both let the COPS Scale's 38 indicators reach the jury without any judicial scrutiny.

The "Considering Offender Probability in Statements" (COPS) Scale organizes 38 indicators into six categories: who the call is about, what the call is about, how the call is made, passive defenses, resistance to dispatcher, and additional guilty indicators. The research base behind those indicators is:

• A 2006 master's thesis built on 100 calls (50 guilty, 50 innocent)
• A 2009 exploratory paper that identified 18 guilt indicators
• A 2017 book that expanded the sample to 200 calls and the indicator count to 38
• No published error rate. No peer review. No independent replication.

The technique has been used in at least 2,500 homicide investigations across 26 states, yet its methodology has never been released for outside testing. That asymmetry is the attack.

Your motion in limine should demand the specific training the witness received, the underlying methodology, and whether the witness is offering a forensic opinion at all. If the State cannot lay the foundation, the "analysis" should not come in under any label, whether expert, lay, or closing argument.

You also need a clean, timestamped, diarized transcript of the 911 audio before you can rebut any of this. The technique operates on "words and phrasing, including how someone requests help," and indicators like whether the caller made "an immediate plea for help for the victim." To counter that, you need to show the jury what the caller actually said, when, and in what order. That's what 911 call transcript analysis gives you: word-level transcripts with citations to the exact second, so every rebuttal point is grounded in the record rather than in the State's characterization.

The same discipline applies to jail call transcription when the State tries to use post-arrest calls as consciousness of guilt.

Your Four-Part Attack: Discovery Demands, Motions in Limine, and Record Preservation

The State's 911 call analysis case looks strongest right up until you ask what the analyst actually did. That question, pressed in discovery and again at the podium, is where this testimony usually comes apart.

Start with a written discovery demand aimed at the training itself. Ask for the instructor's materials, the underlying data set, the indicator list, and every case in which the witness has applied it. The Cornell Law Review's "shadow forensics" analysis explains why this matters: the technique escapes Daubert and Frye review because the method is never disclosed and the founders are never proffered as experts. If the State can't produce the methodology, it can't lay a foundation under any label.

Then file motions in limine that close both doors at once. One motion targets the testimony structure itself: no witness may opine on caller culpability without establishing the training, the data, and the error rate. A second motion pre-empts the workaround the University of Cincinnati Law Review documented, where prosecutors smuggle the technique in through lay testimony or by "guiding juries" through the audio and transcript. Move to bar any walkthrough that applies the COPS Scale's 38 indicators, whether it's called expert opinion, lay observation, or argument.

Your cross has four questions that do most of the work:

• What was the sample size? The 2006 thesis rested on 100 calls; the 2017 book expanded to 200.
• Has any part of this method been peer reviewed?
• What is the published error rate? There isn't one.
• Have you ever been qualified as an expert on this technique?

That last question tends to land hardest. Then walk the witness through the transcript line by line and show the "indicator" the State is relying on is equally consistent with innocence, which is the confirmation-bias problem the Ford School of Public Policy names directly. You can't do that walkthrough without a word-level, speaker-separated transcript you can cite to the second. 911 call transcript analysis produces exactly that, and the same discipline applies to jail call transcription when the State reaches for post-arrest calls as consciousness of guilt.

Preserve everything. A Michigan murder conviction was overturned because 911 call analysis was improperly admitted, and the Illinois Supreme Court agreed to review the Jessica Logan case after ProPublica's investigation documented more than 100 cases across 26 states. Object on the record, proffer what you sought, and get the ruling. Even a loss at trial becomes the issue that wins on post-conviction review.

Build the Grounded Transcript Before the State Narrates Your Client's Call

Harpster's own materials tell prosecutors they can "guide juries through both the audio and transcript of the calls" to "create lasting impressions," as the University of Cincinnati Law Review documents. That guidance only works if the State controls the narrative of what your client said. Your counter is simple: build the record first, word by word, before anyone characterizes it for the jury.

That means a word-level, timestamped, diarized transcript of the 911 audio. The technique operates on phrasing and ordering, indicators like "no immediate plea for help for the victim indicates guilt," drawn from a 2006 thesis built on just 100 calls and a 2017 book that expanded the sample to 200. When you can cite the exact second the caller actually asked for help, the indicator stops being a forensic conclusion and becomes a testable claim you can impeach against the record.

A grounded workflow looks like this:

• Transcribe the call with speaker diarization and word-level timestamps, so every rebuttal point cites a specific second rather than a paraphrase.
• Map each claimed COPS Scale indicator to the actual transcript language, and note where the words contradict it or fit innocence equally well.
• Have an attorney verify the transcript against the audio before it anchors any motion or cross-exam outline.

Evidence Copilot handles this with citation-to-the-second transcript review: diarized 911 transcripts where every AI answer links to the exact moment it came from, always marked for human review, with originals left untouched. The same service page for 911 call transcript analysis covers the full workflow.

Apply the identical discipline to jail calls. Prosecutors routinely package post-arrest calls as consciousness-of-guilt evidence, and the same characterization risk applies: a detective summarizing "what the defendant said" versus the actual words at the actual timestamp. Grounded jail call transcription lets you quote the real sentence, in context, at the second it was spoken.

The stakes are concrete. A Michigan murder conviction was overturned because 911 call analysis was improperly admitted, and the Illinois Supreme Court agreed to review Jessica Logan's case after ProPublica documented 100-plus cases across 26 states using the method. Courts are increasingly willing to reverse when the record shows what the caller actually said. Build that record yourself, to the second, and you impeach the indicator instead of arguing with a characterization.