The Voice of Frida Kahlo? An Acoustic–Phonetic Approach to the Case

This study examines a voice recording attributed to Mexican painter Frida Kahlo, which was discovered in the files of Mexico's National Audio Archives (MNAA). The case gained attention when actress Amparo Garrido claimed that the voice in the recording was hers. Due to the absence of other known recordings of Kahlo's voice, this study tests the following hypotheses: that Garrido is the speaker (H so ) or that she is not (H do ). To address this question, the multivariate likelihood ratio approach was employed to analyze vowel formant frequencies. The system was then calibrated and validated with data provided by the MNAA that reflect the conditions of the relevant population as well as the questioned- and known-speaker recordings. A log 10 likelihood ratio of −2.2 was obtained, providing moderate evidence against Garrido being the speaker. The study provides a clear demonstration of the usefulness of formant analysis in forensic voice comparison.