We studied the transmission bandwidth required for accurate diagnoses when performing realtime fetal tele-ultrasound consultations. The study was divided into three phases. In phase I, three experienced clinicians evaluated the quality of ultrasound images transmitted at various bandwidths (internally looped back within Brisbane) using eight commercially available codecs at random. The two codecs that performed best proceeded to phase 2, in which a realtime video-link of up to 2 Mbit/s was set up between Brisbane and Townsville (1,500 km apart). Testing with a standard video-tape was performed at seven different bandwidths selected at random, with four clinicians (who were blinded to the equipment and bandwidths used). The optimum line rates for transmission were determined, and testing was then performed using these line rates for fetuses with various anomalies (phase 3). The results showed significant differences in performance according to bandwidths used, but not according to observers. Bandwidths were grouped into three levels. At level I (256 kbit/s) the performance was significantly worse than at level II (384, 512 or 768 kbit/s), which was in turn worse than that at level III (1, 1.5 or 2 Mbit/s). However, within each level, performance at one bandwidth was not significantly different from that at the others. The most cost-effective transmission rates therefore appeared to be 384 kbit/s and 1 Mbit/s. Further testing with fetuses affected by various anomalies confirmed that the majority could be diagnosed using a 384 kbit/s link, with slight improvement in evaluation when the bandwidth was increased to 1 Mbit/s.