What Prosecutors Are Doing With 911 Call Analysis, and Why Courts Keep Pushing Back

Prosecutors have used 911 call analysis in over 100 cases across 26 states, and courts are increasingly refusing to let it in. The technique claims word choice and cadence reveal guilt, but the science behind it doesn't hold up to scrutiny.

Here's the problem for you as a defender: the method is everywhere in prosecution training, and it's built on a research base of roughly 200 calls. Harpster's 2006 master's thesis started with 100 calls and 16 factors. A 2009 paper bumped that to 18 indicators. By 2017, the book expanded the sample to 200 calls and claimed 38 indicators under a proprietary "COPS Scale." The indicator count shifted every time the theory was published, which is itself a reliability argument you can put in a motion.

Worse for the prosecution's position, Harpster and co-author Dr. Susan Adams have never released their underlying data, their methodology for obtaining the calls, or anything that would let another researcher replicate the work. As the University of Cincinnati Law Review documents, no one has replicated the findings. That's fatal under Daubert and it's fatal under Frye.

Courts have noticed. Harpster was rejected as an expert in Alabama in 2009. A Nevada judge bounced a detective trained in the method. In Michigan, the conviction of Riley Spitler, a 16-year-old defendant, was overturned because 911 call analysis and other statement-analysis techniques should not have been admitted. Harpster has never testified as a criminal-trial expert. So prosecutors found a workaround.

What they do now is skip expert qualification entirely. Trained prosecutors are encouraged to "guide juries through both the audio and transcript of the calls" and to weave the analysis into lay testimony. When a witness starts "interpreting" what a caller's grammar means, that's the technique smuggling itself in past FRE 702.

The institutional backlash is loud and it's coming from inside the house:

• ProPublica's investigation documented the 100+ cases and prompted the Illinois Supreme Court to agree to review Jessica Logan's conviction
• North Carolina's Office of Indigent Defense Services issued a statewide warning telling attorneys to watch for appearances of 911 call analysis
• Fair and Just Prosecution, a network of elected prosecutors, called on members to review past cases for the technique
• NACDL published a member article framing 911 call analysis as forensic junk science, and NACDL's president called the business "dangerous and insidious"

Even law enforcement is distancing itself. Susan Medina, chief of staff at the Colorado Bureau of Investigation, said the agency views 911 call analysis as "nothing more than a potential investigative lead" and would never advocate for it as a deciding factor in an arrest or conviction. Save that quote. It's impeachment gold against any detective who oversells the technique on the stand.

For your own review workflow, the practical move is to get the 911 audio into a workspace where you can pull a timestamped, diarized transcript and compare the dispatcher's questions against the caller's answers line by line. That's exactly what 911 call transcript analysis is built for, and the same grounded-citation approach applies when you're working through jail call transcription on a co-defendant's statements. Every AI output stays flagged for your review, and every citation points at the exact second of audio it came from.

The takeaway: expect the prosecution to keep trying this. Your job is to recognize it early in discovery, object on FRE 701 grounds when it appears as lay testimony, and have the Daubert motion ready when they try to qualify someone.

The Two Exclusion Frameworks: Daubert/Frye and the Confrontation Clause

Two doors keep 911 call analysis out of the record, and you need to know which one to knock on. The first is evidentiary reliability (Daubert or Frye, depending on your state). The second is the Confrontation Clause, and it turns on whether the call is testimonial.

Start with the reliability gate. The method sits on roughly 200 calls total across two small studies, and the underlying data has never been publicly released. That alone gives you a Daubert motion with real teeth.

Under FRE 702 (and the 2023 amendment), the proponent must show the expert's knowledge will actually help the jury and rests on sufficient facts. 911 call analysis fails every Daubert factor, and the record supports each failure:

• Testability: no one has replicated Harpster's findings, and the data isn't available to try.
• Peer review: the base is a master's thesis, an exploratory paper, and a book, not a peer-reviewed literature.
• Error rate: none published. Ever.
• Standards controlling the technique: the indicator count shifted from 16 (2006) to 18 (2009) to 38 (2017). A method that keeps moving its own goalposts has no controlling standard.
• General acceptance: courts in Alabama (2009) and Nevada rejected it, and a Michigan murder conviction was thrown out because it came in.

In Frye jurisdictions, you argue general acceptance directly: no court has admitted Harpster as an expert, and he has never testified as one. That is a clean, quotable record.

Now the workaround, because this is where defenders actually lose. Prosecutors trained in the method are encouraged to walk juries through the audio and transcript without ever seeking expert qualification. That is 911 call analysis smuggled in as lay opinion under FRE 701. Object the moment a detective or a "trained" civilian starts interpreting word choice, pauses, or grammar. FRE 701 caps lay testimony at what a layperson could perceive, and none of these guilt indicators are perceivable without the training.

The Confrontation Clause is a separate fight. Under Davis v. Washington (2006), statements made during a 911 call in the course of an ongoing emergency are nontestimonial, so Crawford does not bar them. But the line moves: if the caller is describing past events after the emergency ends, you have a strong argument the statement becomes testimonial and requires cross-examination.

One more trap before you play that call for the jury. In State v. Outland (N.J. App. Div. 2019), the court held that introducing a defendant's 911 recording is the equivalent of an assertion of innocence from the witness box, which opened the door to prior-conviction impeachment under N.J.R.E. 806 even though the defendant never testified. Check your jurisdiction's analog before you play it.

When you do need to attack the call's content, get a timestamped, diarized transcript first. Evidence Copilot's 911 call transcript analysis and jail call transcription give you word-level timestamps and citations to the exact second of audio, so your FRE 701 objection can point to the precise moment a witness overstepped.

The Impeachment Trap: State v. Outland and What Happens When You Play the Call

In New Jersey, playing your client's own 911 call at trial can hand the State the impeachment ammunition you spent the whole case trying to keep out. That's the holding in State v. Outland, and it should shape how you weigh the decision to introduce the recording at all.

In Outland, the Appellate Division ruled that playing a 911 recording for the jury is "in effect, equivalent to making an assertion of innocence from the witness box." That characterization opened the door to impeachment with the defendant's prior convictions under N.J.R.E. 806, even though he never testified. The stakes in that case were real: a sixteen-year extended term as a persistent offender, roughly $35,000 stolen, and four prior indictable offenses.

The logic tracks the hearsay exceptions that get 911 calls admitted in the first place. Courts commonly allow them as present sense impressions or excited utterances, and in Outland the defendant's voice sounded "excited and unreflective." The call comes in as your client's own out-of-court statement, which means the State can attack it the same way it would attack testimony. Before you play the audio, work through the questions:

• Does the call actually help on the element you're fighting, or does it just humanize a bad fact?
• What prior convictions does your client have, and are they admissible under your state's Rule 806 analogue?
• Can you get the same content in through a stipulation or a narrower excerpt that doesn't function as an innocence claim?
• If your client never testifies, do you want the jury hearing his voice assert innocence anyway?

The research base behind 911 call analysis makes the decision harder, not easier. Tracy Harpster's method rests on about 200 calls across two studies, and the indicator count has shifted over time, from 16 factors in his 2006 thesis to 18 indicators in 2009 to 38 in his 2017 book. No one has replicated the findings, and the underlying data has never been publicly released, as the University of Cincinnati Law Review documents. When the State offers any interpretation of a call, that record is your objection.

If you do play the recording, listen to it first with a timestamped transcript in hand, then run the dispatcher's questions against the caller's answers and against the police report and bodycam. That comparison work is exactly what 911 call transcript analysis is built for, and it pairs with jail call transcription when the same client's recorded statements show up in custody. Every AI output is marked for your review, and every citation points to the exact moment it came from.

Building the Defense Workflow: Transcript, Compare, Cite

You have the audio file. Now you need the words, in order, with timestamps. Accurate transcription is the foundation of any 911 call analysis. Without a precise transcript, you cannot compare the dispatcher’s questions against the caller’s answers. You also cannot spot the subtle shifts in tone or timing that might undermine a prosecution narrative.

Start by obtaining the raw 911 audio from discovery. Ensure the file is complete and unedited. Then, generate a timestamped transcript that identifies each speaker. This step is critical because prosecutors often rely on guided playback to steer juries toward specific interpretations. You need an objective record to counter that narrative.

Here is a practical workflow for reviewing the call:

• Identify the dispatcher’s questions and note the exact second they are asked.
• Transcribe the caller’s responses word-for-word, paying attention to pauses and hesitations.
• Cross-reference the call timeline against the police report’s initial arrival time.
• Compare the caller’s description of the scene with bodycam footage from the responding officers.
• Flag any discrepancies between the 911 statement and subsequent jail calls.

This process allows you to detect inconsistencies that a quick listen might miss. For example, if the caller describes a "struggle" in the 911 call but the bodycam shows no signs of disturbance, that is a significant defense point. Similarly, if the caller’s tone changes abruptly after a specific question, you can isolate that moment for cross-examination.

Many defense teams struggle with the volume of media in modern cases. A single incident can include multiple bodycam angles, dashcam footage, and several phone calls. Managing these files manually is time-consuming and error-prone. Tools like 911 call transcript analysis help streamline this process by providing accurate, diarized transcripts that you can search and cite. This saves hours of manual review and ensures you do not miss critical details.

If your case includes statements made in custody, consider jail call transcription as well. These calls often contain admissions or inconsistencies with the initial 911 report. By transcribing and analyzing these calls, you can build a stronger defense strategy based on factual discrepancies rather than speculative theories.

Remember, the goal is not to prove guilt or innocence from the audio alone. The goal is to ensure the jury hears the full, unedited context. If the prosecution tries to use 911 call analysis as expert testimony, you have strong grounds to object. Institutional backlash against the technique is growing, with many judges rejecting it as unscientific. Your job is to present the facts clearly and let the jury decide.

By building a rigorous workflow around transcript comparison and cross-referencing, you turn a potentially biased narrative into a defensible factual record. This approach respects the integrity of the evidence while giving you the tools to challenge it effectively.