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

Your client says "hi" to a 911 dispatcher. A detective later testifies that greeting was a guilt indicator, and the mother ends up taking an Alford plea to manslaughter. That's not a hypothetical: it's a documented 2018 Washington case, and it's the kind of reasoning that keeps showing up in your discovery.

911 call analysis is a technique claiming that a caller's word choice, politeness, and cadence can reveal whether they're guilty of the crime they're reporting. It was built by Tracy Harpster, then a Moraine, Ohio police lieutenant, starting with a 2006 master's thesis on 100 homicide calls. A 2009 paper with FBI analyst Susan Adams expanded it to 18 guilt indicators, and their 2017 book grew the framework to 38 indicators across roughly 200 calls, packaged as the proprietary "COPS" scale (Considering Offender Probability in Statements).

Here's what should trouble you: the entire research base is two small studies of about 200 calls total, the underlying data has never been released, and no one has successfully replicated the findings. Harpster has nonetheless consulted on at least 2,500 homicide investigations as of March 2024, and prosecutors and investigators in at least 26 states have used the technique. He also earns "thousands of dollars" per eight-hour training course sold to police departments.

So why does this land on your desk without a Daubert fight attached? Because courts have repeatedly rejected it as expert testimony. Harpster himself was excluded in Alabama in 2009, and a Nevada judge refused a detective trained in the method. His workaround, documented in a University of Cincinnati Law Review analysis, is to coach prosecutors to "guide juries through both the audio and transcript of the calls" as lay testimony, embedding the analysis in walkthroughs and closing arguments where expert-admissibility standards never get triggered.

That's the practical trap. The analysis rarely arrives labeled as analysis. It shows up as:

• A detective's lay testimony that the caller was "too polite" or "didn't ask for help" for an innocent person
• A report or supplemental narrative flagging the call as suspicious based on COPS-scale language
• An interrogation where a detective confronted your client with a call-analysis handout, as happened in a 2012 Fremont, Ohio case that produced a manslaughter plea
• Closing-argument commentary walking the jury through the audio with guilt-indicator framing

The University of Michigan's Ford School policy brief independently concluded the technique lacks robust empirical support, reinforces confirmation bias, and has contributed to wrongful convictions. Michigan's Riley Spitler murder conviction was overturned in part because this analysis shouldn't have been admitted.

Before you can challenge any of it, you need the call itself: authenticated, accurately transcribed with speaker IDs and timestamps, and searchable against everything else in discovery. When the prosecution's theory turns on exactly how a phrase was delivered, a sloppy transcript is a problem. That's the gap Evidence Copilot is built for, giving defense teams word-level, diarized transcripts of 911 and jail call audio with citations to the exact second, so you can see for yourself whether the "indicators" the detective described actually exist in the recording.

The Lay-Testimony End Run Around Daubert

The 2009 Alabama trial is where this all starts to unravel. Harpster himself was rejected as an expert witness there, and a Nevada judge later excluded a detective trained in the same method. Courts have drawn a hard line: 911 call analysis doesn't clear Daubert or Frye. So the technique didn't disappear. It went underground.

Prosecutors figured out they don't need an expert designation. They just need an officer with a story to tell. Harpster's own training materials advise prosecutors to walk juries through both the audio and transcript of the call and to "create lasting impressions of this critical evidence." He's published ten tips for prosecutors on doing exactly this without ever qualifying as an expert. The University of Cincinnati Law Review calls it what it is: the work gets "mixed into in-court statements and lay testimonies to sneak it past the standards for expert testimony."

That's the 2026 defense problem. You can win the Daubert motion and still lose the trial, because the same junk analysis shows up dressed as common sense. Here's what to scrub for when you review officer and dispatcher testimony:

• COPS-scale vocabulary in lay answers: "too polite," "didn't sound panicked enough," or treating "I need help" as a guilt indicator (in one Michigan case, that phrase alone made a caller a suspect).
• Jury "walkthroughs" of the 911 audio and transcript where the officer narrates what the caller "really" meant by word choice or cadence.
• Closing arguments that repeat the same guilt-indicator framing the officer introduced on direct.
• Detective reports or handouts that reference the Harpster method by name, which then get laundered into "investigative experience" on the stand.

Your objection is FRE 701 versus 702. Lay opinion has to come from personal perception and can't rest on "scientific, technical, or other specialized knowledge." When an officer explains why a caller's greeting of "hi" signals guilt (a real example from a 2018 Washington case that ended in an Alford plea), that's specialized knowledge wearing a badge. Object early, and ask for a foundation: what training did this officer receive in statement analysis, and when?

Preserve the record even if you lose the objection. Ask the officer to identify every source of their opinion, name the training course, and confirm they can't point to peer-reviewed support. Remember, the entire research base behind this technique is roughly 200 calls across two studies, with zero successful replications and no public data.

Two practical notes on the evidence itself. Get the call early via subpoena to the custodian of records, and get it transcribed with speaker IDs, timestamps, and notation of pauses and intonation, since accuracy "may become the focal point of a legal dispute." For a working transcript you can cite to the second, 911 call transcript analysis from Evidence Copilot gives you timestamped, diarized text with every AI observation tied to the exact moment it came from.

One more place this shows up: the interrogation room. In a 2012 Fremont, Ohio case, a detective used a 911 call analysis handout to elicit a confession that led to an 11-year manslaughter sentence. If your client was confronted with "indicators" from their own call before confessing, that's a voluntariness and suppression argument grounded in the technique's documented scientific invalidity. Document the handout, the questions, and the timeline. And if the state is building its narrative on hours of recorded calls, jail call transcription keeps the full record searchable and citable when you're ready to confront the theory at prelim.

ctaText: See how grounded citations work on your own case files: book a 20-minute walkthrough at evidencecopilot.com/book socialProofText: Defense teams use Evidence Copilot to keep every claim tied to the exact second of video or page of the PDF, with AI output marked for human review and published pricing at $39/user/mo on the Solo plan.

Building the Daubert/Frye Challenge

When a detective walks into chambers and announces an "expert" opinion about your client's guilt based on how they said "help" during a 911 call, you have a Daubert problem with a confession attached. The technique's entire research base is two studies of 100 calls each, and no one has replicated Harpster's findings. That is the foundation you attack.

Frame the motion around the five Daubert factors and let the documented weaknesses do the work. Under the testability prong, the data has never been released. Under peer review and publication, the methodology is opaque. Under error rate, none has ever been calculated. Under general acceptance, the technique has been rejected by courts, including a 2009 Alabama trial where Harpster himself was kept off the stand. Under standards, the indicators have ballooned from 16 in the 2006 thesis to 18 in the 2009 paper to 38 in the current COPS Scale, a moving target that signals post-hoc refinement rather than validated science.

The strongest precedent paragraph cites cases where courts already found this unreliable. In Michigan, Riley Spitler's murder conviction was overturned because 911 call analysis and other statement-analysis techniques should not have been admitted. In Alabama, the trial court excluded Harpster as an expert. In Nevada, a judge rejected a detective trained in the method. These are not hypothetical concerns. They are rulings on the record.

The harm section converts the Daubert analysis into a constitutional argument. In a 2012 Fremont, Ohio case, a detective used a 911 call analysis handout to elicit a confession; the defendant pled to manslaughter and received an 11-year sentence. In a 2018 Washington case, a mother's greeting of "hi" was treated as a guilt indicator, the death was reclassified as a homicide, and she took an Alford plea to manslaughter. In a Michigan case, a caller was flagged as a suspect for saying "I need help." The University of Michigan STPP policy brief confirms the technique has contributed to wrongful convictions and recommends restricting its use.

Your Frye paragraph tracks the same facts through the general-acceptance lens. A method used in 26 states sounds widespread until you note that the creator earns thousands of dollars per training session and the underlying research base is roughly 200 calls. General acceptance by practitioners who paid for the training is not the same as acceptance by the relevant scientific community, which has produced zero replications.

Close the motion with a specific prayer: exclude any expert testimony applying the COPS Scale or its variants, and preclude lay witnesses from offering guilt conclusions drawn from word choice, politeness, or cadence in the 911 recording. That second request matters because prosecutors are coached to route the analysis through officer testimony and jury walkthroughs of the audio rather than seek expert qualification, which is exactly the end-run the law review warns about.

When you build the evidentiary record, a clean 911 call transcript analysis with speaker IDs, timestamps, and notation of pauses lets you cross-reference the prosecution's "indicators" against the actual audio. If the same discovery includes jail calls, pairing them through a unified jail call transcription workflow keeps every citation grounded in the exact moment of the recording, which is the only foundation that survives cross-examination.

Getting the Call Early and Turning It Into Defense Evidence

The 911 call is often the first piece of evidence you control. Getting it early and transcribed correctly turns a passive recording into an active defense tool.

Do not wait for the prosecutor to play the audio for the first time. Subpoena the custodian of records immediately to secure an authenticated copy. In jurisdictions where a custodian declaration is accepted, use it to streamline the process, but be aware that attorneys disagree on whether this document is self-executing without live testimony. Where possible, seek a stipulation with the opposing counsel to avoid foundation disputes at trial.

Transcription quality is critical because accuracy can become the focal point of a legal dispute. Emergency call audio is often poor quality, filled with dropouts, panic, and slang. You need a professional transcript that captures speaker IDs, precise timestamps, and notated pauses. These details matter when the prosecution’s theory hinges on exactly how a word was stressed or a silence was interpreted.

Once you have a clean transcript, deploy the call affirmatively. Use the recording to impeach witnesses whose trial accounts diverge from the initial report. It is also a powerful tool to refresh a witness's recollection or contradict inaccuracies in police reports. If the case involves interrogation, pay close attention to the context. Documented cases show detectives using 911 call analysis handouts to elicit confessions. For example, in a 2012 case, a detective used such a handout to help secure a plea to manslaughter. If your client confessed after being confronted with these unverified "indicators," that context is a prime target for a motion to suppress based on the technique's scientific invalidity.

To manage this workflow efficiently, consider using a platform like 911 call transcript analysis services that provide timestamped diarized transcripts. This ensures every second of audio is searchable and citable, allowing you to pinpoint inconsistencies between the call and subsequent police statements.

Key steps for leveraging the 911 call:

• Subpoena the custodian of records to secure an authenticated copy early in discovery.
• Obtain a high-quality transcript with speaker IDs and notated pauses to prevent accuracy disputes.
• Use the call to impeach witnesses or refresh recollection when their accounts conflict with the recording.
• Flag interrogation contexts where call-analysis handouts were used to elicit confessions as suppression targets.

For related audio evidence, such as statements made during booking, jail call transcription provides similar benefits. It helps you build a comprehensive timeline of the defendant's statements, ensuring no inconsistencies are missed between the initial 911 call and later interactions with law enforcement. This thorough review allows you to challenge the prosecution's narrative with grounded, citable evidence.

Working the Call in a Grounded Review Workflow

When a prosecutor hands the jury a transcript and walks them through a 911 call line by line, your counter isn't a hunch. It's a workspace where the recording, the report, and the testimony sit side by side, each one citable to the second.

Build the case around a diarized, timestamped transcript of the 911 audio. Speaker labels, word-level timestamps, and one-click renaming that propagates across every segment mean the dispatcher's question and the caller's answer stay attached to the right voice, which matters when the prosecution's theory turns on exactly how something was said. Emergency-call audio is notoriously difficult to transcribe, and accuracy can become the focal point of a legal dispute, as transcription practitioners note.

Then put everything else in the same room:

• The 911 audio transcript, diarized and timestamped
• Bodycam and dashcam footage synced to a single master clock
• Jail calls, transcribed and searchable
• The police report, OCR'd so every page is citable
• Officer and dispatcher testimony, cross-referenced against the recording

That last item is where the work pays off. Because 911 call analysis is rarely admitted as expert evidence, prosecutors increasingly coach officers to embed its claims in lay testimony and jury walkthroughs, effectively sneaking the analysis past expert-testimony standards. When an officer testifies that a caller was "too polite" or that "I need help" signaled guilt, you need the recording and the report in one view, with citations that jump to the exact moment. Evidence Copilot's 911 call transcript analysis and jail call transcription feed that same workspace, and auto-signals compare each new file against earlier evidence so report-versus-call inconsistencies surface as review signals before anyone has to ask.

Every AI output is marked for human review and cites the exact source moment it came from. Published pricing is $39/user/mo on Solo, $99/user/mo on Firm, and $750/attorney/yr for PD offices, with a 7-day full-access trial and a 14-day money-back guarantee. Book a 20-minute walkthrough at evidencecopilot.com/book.