We report direct measurements of changes in diving and movement behavior for 53 goose-beaked whales (Ziphius cavirostris) in relation to experimentally controlled mid-frequency (3–4 kHz) active sonar (MFAS) signals. These signals simulate powerful Navy sources that have been associated with multiple mortal stranding events for this species. We deployed a multi-scale combination of tags to monitor individual whales, including 50 long-duration (weeks), coarse-resolution satellite-transmitting tags and 3 short-duration (hours), high-resolution archival depth, orientation, and acoustic tags. We evaluated behavioral responses during 13 experimental trials (9 MFAS; 4 no-MFAS controls), resulting in 72 exposure events; some individuals were exposed in multiple trials. Whales were exposed at known and modeled horizontal ranges from ~2 to >200 km and from below ambient noise levels to received levels (RLs) up to ~142 dB re: 1μPa (root-mean-square [RMS]). We investigated changes in diving and movement behavior separately, with a suite of metrics, descriptive evaluations, and statistical tests. We observed similar patterns and probabilities of behavioral changes for control trials and the lowest RL conditions (<100 dB). Above 100 dB RLs, increasingly prevalent and consistent responses occurred, including extended deep dives, prolonged periods between deep dives, directed spatial movement away from the source, and cessation of echolocation. Aspects of these cryptic responses typically persisted for hours following exposure but did not result in broad-scale habitat abandonment. Our study builds upon experimental and observational studies conducted on sonar testing ranges and expands our understanding of the response of this species to MFAS in a region where operational sonar use occurs far less commonly than on Navy testing ranges. These data are directly applicable in the conservation and effective management of this sensitive, protected species.