The cognitive evidence spectrum is a four-category framework introduced by Nita Farahany in her 2012 Stanford Law Review article Incriminating Thoughts. It replaces the Fifth Amendment's testimonial/physical binary, derived from Schmerber v. California (1966), with a more granular spectrum that sorts compelled evidence into four types: identifying, automatic, memorialized, and uttered. The framework is applied throughout Week 10 of Farahany's Advanced Topics in AI Law and Policy course.
The four categories
Identifying evidence (least protected)
Physical traits used to establish who someone is, with no mental participation required. The person contributes only their body; nothing from their mind is accessed or expressed. Examples include fingerprints at booking, DNA swabs, mug shots, voice exemplars, and handwriting exemplars. Under Schmerber, this evidence falls outside the Fifth Amendment privilege.
Automatic evidence
Involuntary physiological responses that may reveal something about a person's internal state but occur without the person choosing to disclose anything. Examples include blood pressure during polygraph, galvanic skin response, pupil dilation, fMRI BOLD signal at rest, and EEG-recorded brain activity not tied to a deliberate task. The constitutional treatment is doctrinally unsettled. Most courts have treated automatic evidence as physical, and thus outside the privilege, but per Farahany the analogy to Schmerber's blood-draw is weakening as the data becomes more semantically rich.
Memorialized evidence
Information stored in memory that can be accessed and communicated. The privilege typically engages here because the only way to reach this information is to compel a mental act — accessing stored knowledge and expressing it. Examples include a passcode, a safe's combination, knowledge of where specific documents are hidden, the location of a body, and the identity of co-conspirators. Under Hubbell (2000), this evidence is protected when the government cannot independently establish the foregone-conclusion elements of existence, possession, and authenticity.
Uttered evidence (most protected)
Direct verbal or written testimony — answers to questions, written statements, confessions. This category receives core Fifth Amendment protection.
Application to the Payne vs. Brown circuit split
Farahany uses the spectrum to analyze the 2024 biometric-unlock circuit split. In U.S. v. Payne (9th Cir. 2024), officers forcibly grabbed the defendant's thumb, and the court held there was no Fifth Amendment violation. The spectrum reads this as identifying evidence, the least protected category, on the basis that Payne was physically passive and his thumb was used as an instrument, comparable to fingerprinting at booking. In U.S. v. Brown (D.C. Cir. 2024), an agent ordered the defendant to actively unlock with a thumbprint, and the court held there was a violation. The spectrum reads this as memorialized-becoming-uttered, on the basis that the act of authentication communicates knowledge of the device, control, and that this finger is the credential — a mental act expressed through movement.
Both courts applied the same three-element test of compelled, incriminating, and testimonial conduct. According to Farahany, they measured "testimonial" differently because the testimonial/physical binary does not have the right distinctions built in, and the spectrum makes the disagreement legible by identifying which evidence type each case involved.
Application to AI and neurotechnology
Farahany argues the framework becomes more pressing as authentication interfaces become more seamless. Under the Payne "cognitive exertion" framing, she notes, a Meta EMG band that reads intention before a gesture is even formed would require zero cognitive exertion and thus receive zero Fifth Amendment protection — meaning that, on that logic, the more directly a technology reads a person's mind, the less protection the mind receives. The spectrum instead points toward a rule in which what is being measured (mental content, even if accessed without conscious effort) determines protection, rather than the cognitive effort required to produce it.
Applied to AI conversation logs, the open question raised in Class 10, Farahany suggests that logs reflecting the process of cognition rather than communication of a completed thought are closer to memorialized-becoming-uttered than to identifying or automatic evidence, and on the spectrum's logic should receive substantial Fifth Amendment protection. The doctrine has not yet reached this conclusion.
Relationships
- introduced-by: Nita Farahany (2012 Stanford Law Review)
- load-bearing-in: Inside My Advanced Topics Class 10.2: Two Courts, One Test, One Thumb (Farahany, March 2026), Inside My Advanced Topics Class 10.3: When the Interface Is Neural (Farahany, April 2026)
- depends-on: Foregone Conclusion Doctrine
- related: Cognitive Liberty