A tilt anomaly preceded a pair of earthquakes (ML = 4.2, origin time 0014 UTC, and ML = 3.9, origin time 0018 UTC, both on 29 August 1978) on the Calaveras Fault near San Jose, California. These earthquakes occurred at hypocentral depths of 8.5 and 9.0 km, respectively, and were located 6.7 and 5.2 km northwest of the Mt. Hamilton tiltmeter site. The anomaly is similar in shape and time scale to signals observed on other tiltmeters at the times of recorded surface creep events. The anomaly began approximately 40 hr before the earthquake pair and consisted of gradual down-to-the-east tilting followed by rapid tilting down-to-the-north-northeast at a rate of μrad/hr. This was followed by 1 hr of rapid down-to-the-east tilting amounting to 1.5 μrad. The maximum peak tilt of 10.6 μrad down-to-the-northeast was followed by gradual decelerating tilting down-to-the-southwest constituting partial recovery. An anomaly of nearly identical form, but smaller in amplitude and duration, preceded an ML = 2.2 aftershock on 5 September 1978. Other nearby earthquakes as large as ML = 4.7 have occurred without accompanying creep-like signals. A similar, but a much smaller (0.74 μrad) creep-event-like signal preceded an ML = 3.5 earthquake with epicenter 3 km east of the Black Mountain tiltmeter site. In general, however, short-term tilt anomalies such as these are not observed to precede local earthquakes within the central California tiltmeter network. The tilt signal preceding the 29 August earthquake pair may be interpreted in terms of a model of a propagating creep event, at depth, associated with seismic failure at a “stuck” patch on the fault. However, the data are not adequate to constrain the model sufficiently to constitute a test of the hypothesis.