What Prosecutors Are Doing With 911 Call Analysis

Prosecutors have turned 911 call analysis into a courtroom weapon, and they are not waiting for a judge to bless it as science. The technique, created by Moraine, Ohio police lieutenant Tracy Harpster in 2006, claims to spot guilt through behavioral cues in a caller's words, and prosecutors and investigators in at least 26 states have used it. Harpster himself has consulted on more than 2,500 homicide investigations as of March 2024, according to a University of Cincinnati Law Review analysis of the method.

The engine behind it is the COPS Scale, a checklist of guilt and innocence indicators that has grown from 16 items in Harpster's 2006 master's thesis to 38 today, organized into six categories. The indicators are exactly as subjective as they sound. Saying "Huh?" or "I don't know" counts as guilt. So does an awkward phrase or a "Defendant Mentality." An immediate plea for help for the victim signals innocence; its absence signals guilt. The method has never been validated by independent peer-reviewed research, and Harpster has never testified as an expert in a criminal trial despite creating it.

That gap is the defense opening. Under the five-factor Daubert test, judges weigh testability, peer review and publication, known error rate, controlling standards, and acceptance in the relevant scientific community. The COPS Scale fails on each: no independent review, no disclosed error rate, no standardized methodology.

Rather than qualifying an expert, prosecutors trained in the method guide juries through the audio and transcript themselves, smuggling expert conclusions into lay testimony without ever triggering a Daubert hearing. The law review describes Harpster's work as having been "mixed into in-court statements and lay testimonies to sneak it past the standards for expert testimony." Watch for these patterns:

• A prosecutor or detective narrating a caller's tone, pauses, or word choice as evidence of guilt.
• Behavioral interpretation offered by a lay witness with no foundation in psychology or linguistics.
• COPS Scale vocabulary ("defendant mentality," "guilt indicator") surfacing in testimony without an expert on the stand.
• A 911 call played for the jury with running commentary instead of admitted expert analysis.

Object the moment interpretation starts, and demand the gatekeeping inquiry the prosecution is trying to avoid. The stakes are not hypothetical. The technique was rejected as expert testimony in Alabama in 2009 and in Nevada, admitted in the Michigan Riley Spitler case only to be overturned on appeal, and in a 2012 Fremont, Ohio case it helped produce an 11-year prison sentence after a detective used it to elicit a confession.

Before you can attack the interpretation, you need the record itself. Defense teams working 911 audio alongside bodycam, reports, and jail call transcription need every quote tied to the exact moment it came from, which is what 911 call transcript analysis in a grounded review workspace is built to give you.

How to Attack 911 Call Analysis on Daubert Grounds

Under Daubert, the judge must weigh five reliability factors before any expert testimony reaches the jury. The COPS Scale fails every one of them, and the University of Cincinnati Law Review analysis documents exactly why.

Testability. A reliable method makes predictions you can verify against known outcomes. The COPS Scale doesn't do that in any controlled way: its guilt and innocence indicators have never been tested against blinded samples by independent researchers. What you have instead is a master's thesis built on 100 calls, later expanded to 200, authored by the technique's own creator.

Peer review and publication. The method has never been validated by independent peer-reviewed research. Its publication record is a master's thesis, an article in the FBI Law Enforcement Bulletin, and a practitioner book. None of that is independent scientific validation.

Error rate. No error rate has ever been disclosed for the COPS Scale. None. That alone is fatal under Daubert in most jurisdictions.

The remaining two factors round it out:

• Standards controlling operation. The indicator set has shifted from 16 in the original thesis to 18 in the 2009 paper to 38 today, across 6 categories. A method whose own criteria keep changing has no fixed protocol an analyst must follow.
• Widespread acceptance. The technique has spread to prosecutors and investigators in at least 26 states, but adoption by law enforcement isn't acceptance by the relevant scientific community. Harpster himself has consulted on 2,500+ homicide investigations yet has never testified as an expert in a criminal trial.

When the court refuses to seat an expert, watch for the workaround. Prosecutors trained in the method sometimes walk juries through the 911 audio themselves, offering behavioral interpretation as though it were common sense. Object immediately: that's expert opinion dressed as lay testimony, and it deserves the same Daubert scrutiny. Demand a hearing outside the jury's presence.

Build your objection on the record early. File a motion in limine, and if the prosecutor signals intent to "guide" the jury through the call, request a pretrial ruling. Courts have split: Alabama (2009) and Nevada rejected the technique as expert testimony, while a Michigan court admitted it in Riley Spitler before that conviction was overturned on appeal.

For the underlying audio itself, get a clean, timestamped transcript before you argue anything about what the caller said or meant. 911 call transcript analysis for your own review purposes keeps every claim tied to the exact second of audio, which is what you need when the state's version of the call diverges from the recording. The same discipline applies to jail call transcription when the prosecution leans on other recorded statements.

Prepared defense teams build their Daubert motion on the actual audio, not the state's summary of it. Book a 20-minute walkthrough to see how grounded citations into the exact source moment support your reliability challenge.

The Workflow Problem: Reviewing 911 Audio Alongside Everything Else

A 911 call rarely arrives alone. It lands in the discovery pile next to bodycam footage, interrogation video, jail calls, and a stack of police reports, and someone on your team has to make sense of all of it before the preliminary hearing.

The volume is the first problem. A single case can run forty hours of video and thousands of pages of reports. The 911 audio is often the shortest piece, but it carries outsized weight because prosecutors increasingly use it as a behavioral exhibit, not just a narrative one. Detective Tracy Harpster's method has spread to investigators in at least 26 states, and Harpster has consulted on more than 2,500 homicide investigations. When the state is going to argue your client's first word on the call signals guilt, you need to be able to pull that word, its timestamp, and the call-taker's preceding question in seconds.

The second problem is the standard. As Clio's transcription guidance draws the line, AI-generated transcripts serve internal case preparation, surfacing contradictions and building chronologies, while certified court reporters remain the standard for evidentiary transcripts that go to the jury. That distinction matters when you sit down to plan your workflow: the diarized, searchable transcript you use to prep cross-examination is not the same artifact you mark as an exhibit.

A practical workflow looks like this:

• Ingest the 911 audio alongside bodycam, dashcam, and interrogation footage so timestamps align across files.
• Generate a speaker-labeled transcript with word-level timing so you can jump to any phrase the state quotes.
• Pull every call-taker question into a side column, since the call follows a standardized script covering location, callback number, and nature of the emergency.
• Compare the call transcript against the police report narrative and any jail call statements to surface inconsistencies before the state does.
• Flag behavioral-interpretation language in the prosecutor's filings so you can object before it reaches the jury.

That last step is where the workflow pays off. If a detective or prosecutor has been trained in the COPS Scale and starts guiding the jury through "indicators" in the call, you want the exact transcript window, the call-taker's prior question, and the cited page of the police report on one screen. A workspace built around grounded citations lets you build the Daubert challenge in the time you used to spend hunting for the right file.

The same pipeline covers the jail calls that usually arrive in the same discovery dump. Jail call transcription follows the same logic: diarized speakers, searchable text, and every quote linked back to the audio moment it came from. And when the 911 audio itself becomes the focus of a motion or a cross, 911 call transcript analysis keeps every claim tied to a timestamp a judge or juror can verify.

The point is not to replace attorney judgment. It is to make the underlying audio reviewable at the speed the state is already moving.

Building Your 911 Call Review Checklist

When you pull a 911 call from discovery, you are not just listening to audio. You are auditing a standardized script where call-takers ask specific questions about location, nature of the emergency, and caller details. This structure is documented in federal guidance on calling 911. Understanding this baseline helps you spot where a caller’s phrasing deviates from the expected flow.

Prosecutors often lean on behavioral indicators to suggest guilt. However, a University of Cincinnati Law Review analysis notes that the COPS Scale, which includes 38 guilt and innocence indicators, has never been independently validated. It lacks peer review and established error rates. You need to treat these interpretations with skepticism, especially when they are used to elicit confessions.

Your review process should focus on cross-referencing the audio against the police report and other evidence. Start by transcribing the call to create a searchable text file. This allows you to isolate specific phrases, such as a caller saying "I don't know" or making unintelligible comments. These are flagged as guilt indicators in the COPS Scale, but they may simply reflect stress or poor audio quality.

To build a reliable timeline, compare the call’s timestamp against the officer’s arrival time and bodycam footage. Look for discrepancies in how events are described. If the report claims the caller was calm but the audio shows panic, that inconsistency matters. You are looking for gaps between the recorded reality and the narrative presented by law enforcement.

Consider the context of the call-taker’s instructions. Callers are often coached through first aid or told not to hang up until instructed. This coaching can shape the caller’s responses in ways that look suspicious to an untrained ear. A defense attorney must separate the caller’s genuine reaction from the procedural prompts given by the dispatcher.

For internal case preparation, AI transcription tools can help surface these contradictions quickly. They convert audio into speaker-labeled, timestamped text, allowing you to build chronologies without listening to the same file repeatedly. Remember that certified court reporters remain the standard for official court records. AI transcription is for your internal analysis, not for the evidentiary record.

Use this workflow to prepare for suppression motions or cross-examination. If you suspect the prosecutor is using 911 call analysis as a lay witness tactic, object to it. Demand that any expert testimony meet Daubert standards. You can review specific 911 call transcript analysis workflows to see how teams handle these files. Similarly, jail call transcription can reveal statements made after the initial 911 contact.

Keep your review focused on the facts. Do not let unvalidated behavioral theories dictate your strategy. The audio is a raw data point. Your job is to interpret it accurately within the context of the entire case.

< ul class="blog-list">

If you need a structured workspace to manage these files, consider tools that provide grounded citations and human review. Evidence Copilot offers a way to review this evidence with every AI answer tied to the exact source moment. You can start by booking a walkthrough to see how the workflow fits your practice.