[Paper Review] Effectiveness of the GT200 Molecular Detector: A Double-Blind Test

W. Luis Mochán, A. Ramı́rez-Solı́s|arXiv (Cornell University)|Jan 17, 2013

Misinformation and Its Impacts3 citations

TL;DR

This double-blind study tested the GT200 molecular detector, marketed as a remote substance identifier for drugs and explosives, by hiding amphetamine pills and bullets in one of eight identical boxes. The device failed in 17 out of 20 trials, with results statistically indistinguishable from random chance (p = 1/8), conclusively demonstrating it is ineffective and likely manipulated by user expectation.

ABSTRACT

The GT200 is a device that has been extensively used by the Mexican armed forces to remotely detect and identify substances such as drugs and explosives. A double blind experiment has been performed to test its effectivity. In seventeen out of twenty attempts, the GT200 failed in the hands of certified operators to find more than 1600 amphetamine pills and four bullets hidden in a randomly chosen cardboard box out of eight identical boxes distributed within a 90m$ imes$20m ballroom. This result is compatible with the 1/8 probability expected for a completely ineffectual device, and is incompatible with even a moderately effective working one.

Motivation & Objective

• To rigorously test the GT200 molecular detector’s claimed ability to detect drugs and explosives remotely under controlled conditions.
• To evaluate whether the device performs better than random chance when operators are unaware of target locations.
• To provide scientific evidence to challenge the use of the GT200 in law enforcement and judicial proceedings.
• To address growing concerns about human rights violations linked to false positives from the device.
• To respond to calls from the Mexican Senate and human rights bodies for independent scientific validation of the GT200.

Proposed method

• Conducted a double-blind experiment at the Mexican Academy of Sciences using a 90m × 20m ballroom with eight identical cardboard boxes.
• Hidden 1600 amphetamine pills and four bullets in one randomly selected box out of eight, unknown to the GT200 operators.
• Used certified military operators trained to use the GT200, who were unaware of the true location of the hidden substances.
• Performed 20 independent trials with the device, recording each operator’s detection claim.
• Analyzed results using binomial probability distributions to compare performance against random chance (p = 1/8) and moderate efficacy (p = 0.5 or 0.8).
• Included a control condition where operators flipped coins to simulate random responses, which performed worse than the GT200.

Experimental results

Research questions

• RQ1 Does the GT200 detect hidden drugs and explosives with accuracy significantly above random chance in a double-blind setting?
• RQ2 Is the device’s performance consistent with a completely ineffective instrument that functions randomly?
• RQ3 Can the GT200’s results be explained by operator bias or subconscious cueing rather than actual detection capability?
• RQ4 What is the statistical likelihood that the observed detection rate (3/20) could occur if the device were truly effective?
• RQ5 Does the device’s performance change when operators are unaware of target locations versus when they are informed?

Key findings

• The GT200 achieved only 3 successful detections out of 20 trials, a result statistically indistinguishable from random chance (p = 1/8), with a likelihood of 0.95 for complete ineffectiveness.
• The probability of obtaining 3 hits with a moderately effective detector (p = 0.5) is only 0.1%, and for p = 0.8, it is less than 10−9, making such efficacy values highly implausible.
• The device’s performance was fully consistent with a random selection process, as confirmed by binomial distribution analysis and likelihood functions.
• Even a control condition using coin flips produced worse results (1/20 hits), underscoring that the GT200’s performance was not worse than chance.
• When operators were informed of the target location, the GT200 correctly identified it, indicating performance depends on user knowledge, not detection capability.
• The study concludes the GT200 is fundamentally ineffective and likely manipulated by user expectation, rendering it scientifically and legally unsound as a forensic tool.