Abstract: A temperament trait is generally defined as individual differences in behaviour that are present early in life and relatively stable across situations and over time. The aim of this study was to test the existence of a trait <<gregariousness>> in horses, by testing the stability across situations and over time of the responses to different social events. Sixty-six Welsh ponies and 44 Anglo-Arab horses were successively tested at 8 months and 1.5 years of age. Among them, 33 ponies and 21 horses were also tested at 2.5 years of age. They were submitted to four test situations: isolation and separation from, attraction towards and passing conspecifics. We carried out the analysis using each of four test groups as a unit (e.g. 33 Welsh ponies born in 2001, tested in isolation). Isolation and separation stood out as tests that showed a high consistency within test, across tests and across time. The most interesting behavioural parameter was the frequency of neighing, which was well correlated with other parameters measured in the same tests, such as defecation, locomotion and vigilance, as well as across the 3 years (e.g. for separation test: 0.41 < R < 0.61). Therefore, the behaviour of neighing observed in separation or isolation tests as early as 8 months of age appears to be a good indicator of similar behaviour in similar situations later in life, but also of other behaviours which can render the horse difficult to use. No parameter recorded during the attraction test presented stability across situations and time: the reactions to this test were not the expression of a stable characteristic of the individual and did not reflect the same characteristic as the three other tests. Of the different parameters recorded during the passing conspecifics test, the time to cross the arrival line near conspecifics showed good stability across years (0.35 < R < 0.68). This parameter was also correlated with many others recorded during the same test, and also, to a certain extent, to the frequency of neighing in the isolation and separation tests. This stability across responses expressed in various social contexts, and this stability over time, from 8 months to 2.5 years of age suggest the existence of a trait of gregariousness in the horse. From a practical point of view, that means it is possible to estimate the level of gregariousness of a horse as early as 8 months of age. Furthermore, additional analysis shows that gregariousness decreases with age.