The Junk Science Hiding in Lay Testimony

A detective takes the stand and tells the jury your client sounded "guilty" on a 911 call. She never names the method. She never cites a study. She just describes what she heard. That is the whole game.

911 call analysis is a technique claiming to identify guilt and deception from the word choice, cadence, and grammar of 911 callers. It was developed chiefly by former Ohio detective Tracy Harpster, who claims more than 1,500 homicides credited to the method. Departments pay up to $3,500 for an 8-hour training program, often using taxpayer funds.

The science behind it is thin. The original study examined only 100 calls and was intended as exploratory research; one co-author later said the findings were not ready for real-world application. At least eight subsequent studies have failed to replicate or validate the method. Researchers who tried to corroborate Harpster's claims have failed, according to a ProPublica investigation documenting more than 100 cases across 26 states where law enforcement used the technique.

Here is the trick: the technique's underlying methods are never disclosed and its founders are never proffered as experts. Instead, 911 operators or police officers testify about caller culpability based on training they attended, so the technique itself largely escapes Daubert or Frye challenge. Prosecutors acknowledge the method "will not hold up to scientific scrutiny" but introduce it indirectly through lay testimony from trained detectives and dispatchers.

Cornell Law Review calls this pattern "shadow forensics." It is particularly pernicious because it operates in the shadows, unlike traditional junk forensics, and follows the same trajectory as bitemark evidence, hair microscopy, arson investigation, and toolmark comparison. Faulty forensic disciplines, including misleading testimony, are the second most common factor in wrongful convictions.

Dr. Jessica Salerno's research shows what investigators interpret as "guilt" is often the absence of expected emotion. Officers required HIGH emotional displays to reduce suspicion. Studies using real 911 calls with known outcomes found low emotion associated with guilt regardless of whether callers were later convicted or exonerated.

When you review the discovery dump, watch for these red flags:

• Officer testimony about caller "indicators of guilt," word choice, or cadence
• Detectives or dispatchers who attended Harpster-style training
• Prosecutors describing the call as showing "deception" or "guilt"
• Expert reports that never name the methodology

North Carolina's Office of Indigent Defense Services has already issued a statewide warning to attorneys to watch for these appearances. NACDL president Nellie King called the business of 911 call analysis "dangerous and insidious."

Because the technique largely escapes Daubert/Frye challenge when introduced via trained officers' lay testimony, your attack has to target the lay/expert line, not just the underlying science. Dr. Salerno specifically recommends defense attorneys "learn to identify and challenge attempts to introduce 911 call analysis through lay testimony."

A grounded review workspace helps here. When you upload the 911 audio alongside the transcript, word-level timestamps and diarized transcripts let you compare what the transcript says against what the caller actually sounded like. That transcript-vs-audio comparison is where impeachment material lives: tone, hesitation, and background sounds that a written transcript cannot capture.

If you are reviewing a case where the prosecution is leaning on a 911 call, start with the audio. Run a 911 call transcript analysis that pairs the transcript against the original recording, and check every officer statement about caller behavior against the actual timestamps. The shadow only stays a shadow if nobody shines a light on it.

The Failed-Replication Record You Can Use in Briefing

The original study behind 911 call analysis examined only 100 calls and was explicitly intended as exploratory research. A co-author even stated the findings were not ready for real-world application. Yet police departments continue to use this method in courtrooms across 26 states.

This gap between the science and the courtroom creates a powerful opportunity for defense attorneys. You can deploy the failed-replication record affirmatively in admissibility motions and cross-examination. The data shows that at least eight subsequent studies failed to replicate or validate the technique’s claims.

Dr. Jessica Salerno’s research further undermines the core premise of the analysis. Her work demonstrates that what investigators interpret as "guilt" is often just the absence of expected emotion. Officers actually required high emotional displays to reduce suspicion, meaning low emotion is not a reliable indicator of deception. This empirical contradiction is a critical point to highlight when challenging an officer’s testimony.

When building your brief, focus on these specific data points:

• The original study was exploratory and limited to a sample of only 100 calls.
• At least eight independent replication attempts have failed to corroborate the method's validity.
• Research by Dr. Salerno proves the "low emotion equals guilt" heuristic is scientifically backwards.

Use this record to attack the lay-testimony workaround. Because the technique often enters evidence through trained officers rather than experts, it escapes traditional Daubert scrutiny. You must challenge the officer’s qualifications directly by presenting this replication failure data. Show the court that the method lacks the scientific consensus required for reliable evidence.

For practical workflow support, consider using a tool like 911 call transcript analysis to compare the transcript against the original audio. This helps you identify tone-vs-content mismatches that support your argument. If the case involves jail calls, jail call transcription can also reveal inconsistencies in the defendant's statements.

Remember that the technique’s creator claims it has led to over 1,500 homicide convictions. However, the scientific community disagrees with this application. Your role is to ensure the jury is not misled by unverified behavioral claims. Ground your objections in the documented failure of replication studies.

The defense bar is increasingly recognizing this technique as "shadow forensics." By citing the specific number of failed replications, you shift the burden back to the prosecution. They must prove the method is scientifically valid, a standard it has yet to meet. This approach strengthens your motion to exclude or your cross-examination strategy.

Stay focused on the empirical record. The numbers speak for themselves. A method that cannot be replicated is not a reliable forensic tool. Use this knowledge to protect your client’s rights and ensure a fair trial.

Hearsay and the Nelson Playbook for Excluding 911 Evidence

When the prosecution tries to slip a 911 caller's statements into evidence through an officer's mouth, U.S. v. Nelson (6th Cir. 2013) gives defense counsel a clean exclusion route. The Sixth Circuit reversed a conviction where an officer relayed what an anonymous 911 caller reported, even though other evidence supported the verdict. The hearsay went to the crux of the case, so sufficiency did not save it.

The court's reasoning matters for motion drafting. A limiting instruction did not cure the prejudice because the jury could not realistically ignore the caller's accusation once it was in evidence. The "background information" workaround also failed: the officer's state of mind was never at issue, so the statement could not be reframed as context for police action. What would have survived, the court suggested, was a less detailed formulation that did not transmit the caller's description of the defendant.

That gives you specific officer phrasing to object to. Watch for testimony like "the dispatcher told me the caller said the defendant was the shooter," "the 911 caller identified my client by name," or "the caller reported seeing the defendant run from the scene." Each of those transmits the caller's out-of-court statement about the ultimate issue. The safer alternative, which the Nelson court indicated would pass, is something like "police received a 911 call reporting a shooting at this address," stripped of the caller's identification or description.

For 911 call analysis work, the practical implication is that the same audio file often generates two distinct evidentiary problems: what the caller actually said (hearsay under Nelson) and what an officer claims the caller's behavior reveals (the shadow-forensics pattern documented by the Cornell Law Review). Defenders running 911 call transcript analysis should flag both layers during review. The transcript alone will not catch the officer's paraphrase at trial; you need the original audio with word-level timestamps to identify where the officer's testimony departs from what the caller actually said, and to build the contemporaneous motion record before the limiting-instruction question is preserved for appeal.

A few drafting points worth carrying into the motion:

• Frame the objection around the ultimate issue, not the form of the question. Nelson turned on what the statement proved, not how the officer was asked.
• Argue prejudice from the identity transmission. The closer the caller's words mirror the charge, the harder a limiting instruction cures the error.
• Preserve the record by requesting a mistrial or strike the moment the officer paraphrases the caller, so the appellate court has a clean ruling to review.
• Pair the Nelson motion with a Daubert/Frye challenge if the officer also offers behavioral inferences from the call. The same review pass that surfaces hearsay phrasing often surfaces the "indicators of guilt" testimony that eight failed replication studies have failed to validate.

If the prosecutor drops the officer and plays the recording directly, the hearsay fight shifts to Confrontation and the excited-utterance or present-sense-impression exceptions. But when the officer is the conduit, Nelson is your authority, and the motion should be ready before the officer opens his mouth.

Transcript vs. Audio: Your Impeachment Workflow

A transcript alone will not tell you that the caller laughed while describing a body, or that the "frantic" witness paused eleven seconds before naming a suspect. Transcripts miss tone, hesitation, and background noise, and they carry transcription errors that opposing counsel will exploit if you never check the audio. That gap is where your impeachment lives.

Pull both files for every call. Read the transcript cold, then listen to the original recording with the text in front of you, marking every place the two diverge.

• Tone-vs-content mismatches: calm speech, laughter, or flat affect against a transcript claiming panic.
• Timing discrepancies: delays between waking, discovering, and dialing that the state treats as consciousness of guilt.
• Transcription errors: dropped words, misattributed speakers, and omitted background voices.
• Officer gloss: any testimony about caller "indicators of guilt," word choice, or cadence.

That last item deserves its own screening habit. 911 call analysis, the technique that claims guilt can be read from a caller's grammar and emotion, has appeared in more than 100 cases across 26 states, according to ProPublica's nationwide review. At least eight replication studies have failed to validate it, as the Innocence Project has documented, and the original study looked at just 100 calls as exploratory research.

Because prosecutors introduce the method through lay testimony from trained detectives and dispatchers, it largely escapes Daubert and Frye scrutiny. Cornell Law Review calls this pattern "shadow forensics", and Dr. Jessica Salerno recommends defenders learn to identify and challenge it when it arrives through lay testimony. So treat any report language about caller demeanor, cadence, or "indicators of guilt" as a flag, then hunt the audio for what the officer actually heard.

Word-level timestamps and diarized transcripts make that comparison mechanical. When a report says the caller was evasive, you jump to the second, listen, and quote the real exchange. A grounded workspace like 911 call transcript analysis keeps the transcript and the original recording side by side, with citations that seek the exact moment, so every impeachment point traces back to the source instead of your memory. The same discipline applies to recorded custodial calls, where jail call transcription with speaker labels lets you match an officer's paraphrase against what was actually said. Every AI output stays marked for human review, and the originals never change.

Two objections to preserve while you work. First, hearsay: in U.S. v. Nelson, the Sixth Circuit reversed a conviction because police relayed an anonymous 911 caller's report going to the crux of the case, and a limiting instruction did not cure the prejudice. Second, foundation: if the state's witness describes caller behavior as guilt, make them show the training, the method, and the validation record.

Book a 20-minute walkthrough at evidencecopilot.com/book, or start on the Solo plan at $39/user/mo with 15 media-processing hours included.