The evidence presented in two recent California “voiceprint” cases is reviewed. In one, the defendant, Nehemiah Jackson, was accused of making a telephone call to the Riverside police station as a result of which two policemen were murdered. The call was recorded on a 24‐hour tape recorder with a very low signal/noise ratio and an apparent upper frequency limit of about 2300 Hz. Copies of the recordings will be played, and the spectrograms which were used for comparing the voices of the accused and the unknown caller will be displayed. The case ended in a hung jury. In the other case the defendant, Stephen Chapter, was accused of making a bomb threat. The Court found the defendant not guilty, and the evidence “not reliable in this particular case due to: (a) Mistakes and errors in preparation of the spectrograms used in making the identification. (b) Failure to ascertain the existence of such errors. (c) Demonstrated listening errors in court while under cross‐examination. (d) Tentative misidentification of the court ordered exemplar. (e) Failure to maintain adequate records/logs during conducting of tests.” These two cases raise the issue of how to extrapolate from laboratory studies to forensic situations.