Among the most frequently cited reports in the science of earthquake prediction is that by Fraser-Smith et al. (1990) and Bernardi et al. (1991). They found anomalous enhancement of magnetic-field noise levels prior to the 18 October 1989 Loma Prieta earthquake in the ultra-low-frequency range (0.0110-10.001 Hz) from a ground-based sensor at Corralitos, CA, just 7 km from the earthquake epicenter. In this analysis, we re-examine all of the available Corralitos data (21 months from January 1989 to October 1990) and the logbook kept during this extended operational period. We also examine 1.0-Hz (1-s) data collected from Japan, 0.0167-Hz (1-min) data collected from the Fresno, CA magnetic observatory, and the global Kp magnetic-activity index. The Japanese data are of particular importance since their acquisition rate is sufficient to allow direct comparison with the lower-frequency bands of the Corralitos data. We identify numerous problems in the Corralitos data, evident from both straightforward examination of the Corralitos data on their own and by comparison with the Japanese and Fresno data sets. The most notable problems are changes in the baseline noise levels occurring during both the reported precursory period and at other times long before and after the earthquake. We conclude that the reported anomalous magnetic noise identified by Fraser-Smith et al. and Bernardi et al. is not related to the Loma Prieta earthquake but is an artifact of sensor-system malfunction. ?? 2008 Elsevier B.V.