Because conventional observations over the oceans are not available, especially during tropical cyclones, multi-spectral atmospheric motion vectors (AMVs) estimated from geostationary satellites are routinely assimilated in the numerical weather prediction models at different operational centres across the globe. The derived AMVs are generally validated with radiosonde observations available over land at synoptic hours; however, over the ocean there is a limited scope to assess the quality of AMVs. Over ocean, AMVs can be validated with radiosonde data available from opportunistic ships or using dropsonde data available from aircrafts. In this study, the accuracy of the AMVs derived from the geostationary satellites Kalpana-1 and Meteosat-7 is evaluated over the oceanic region. Radiosonde data available from a ship cruise held in the Bay of Bengal during the period 09 July-08 August 2012 and from the Cal/Val site situated at Kavaratti Island (72.62°E, 10.57°N) in the southern Indian Ocean are used to assess the AMV accuracy. In this study, 83 radiosonde profiles are used to validate the Kalpana-1 AMVs, to allow a better understanding of AMV errors over the Indian Ocean. The RMSVD of Kalpana-1 AMVs for the high-, mid- and low-levels are found to be 7.9, 9.4 and 5.3 m s-1, respectively, while the corresponding RMSVD for Meteosat-7 AMVs are 9.1, 5.5 and 3.7 m s-1. A similar accuracy is observed when the AMVs are validated against the NCEP analyses collocated with the nearest radiosonde locations. The high RMSVD and bias for Kalpana-1 AMVs at the mid-level and Meteosat-7 AMVs at the high-level are associated with the limitation of satellite winds to resolve the upper-level easterly jet in conjunction with errors in the height assignment. This study could help the numerical modellers to assign appropriate observation error over this region during the assimilation of AMVs into the NWP models.