The aim of this study was to determine the most suitable and practical definition of spatial and temporal proximities between grouped-housed horses by comparing GPS position data. In total, 43 horses held in a "HIT Active Stable (R)" were equipped with GPS loggers. Different definitions of spatial (< 1, 2, 3, 4, 5 m) and temporal proximity (< 10, 30, 60, 90, 120 s interruptions between recording points) were investigated for morning pasture periods of 30 days analyzing proximity lengths, number of proximities as well as the network parameters density and fragmentation. With increasing spatial proximity definition, the duration of proximities rose from 9 % (4 compared to 5 m) to 76 % (1 compared to 2 m). The number of proximities also increased accordingly. Ascending temporal proximity definitions provided a lower number of proximities by 13-21 % (10 s to 30 s) depending on the spatial definition and the proximities became longer. Further, with higher spatial proximity definition, the network density ascended and fragmentation decreased, which shows that more horses had proximities and group formation occurred. We identified a proximity definition of 3 m for at least 60 s as the most useful solution because of a decreasing change in proximity length and number due to biological and technical restrictions. Further, it is suspected that this approach helps to better characterize the group formation of horses kept in groups.