In a study of individual variation in the behaviour of tethered sows, records were made of 76 Large White × Landrace sows in their first pregnancy (gilts) tethered in stalls. Scan sampling was carried out from 07.00 to 17.00 h on 2-3 consecutive days every 5 weeks, producing sample sizes at three stages of the pregnancy (Periods 1, 2 and 3) of 48, 74 and 39, respectively. Sows were allocated to stalls at random; the influence of sows in adjacent stalls was analysed by correlation and regression analysis. Time spent standing by gilts in Period 1 correlated with that by their neighbours (n = 48, rs = 0.35, P < 0.05). Total time spent in repetitive behaviour (T; per cent of observations) also showed an association with neighbours' behaviour: it correlated most strongly with neighbours' chain chewing and manipulation (C; per cent of observations) (rs = 0.39, P < 0.01). This association was clearest in gilts on food allowances (F; kg day-1) of < 2 kg day-1 (n = 16, rs = 0.67, P < 0.01). It declined in Periods 2 and 3, which suggested that correlations were mostly caused by an influence of neighbours on the behaviour of newly tethered gilts. Regression analysis showed that food allowance and neighbours' behaviour together accounted for 50% of the observed variation and produced the equation loge T = 5.49 ( ± 0.37 ) - 3.00 ( ± 0.25 ) loge F + 0.18 ( ± 0.06 ) loge C ( RMS= 0.868, r2 = 50.4%. It seems likely that the behaviour of neighbours was an important cause of stress of newly tethered gilts. The results have implications for the design of experiments on housing and provide further evidence that tether stalls are not a suitable system of housing for pregnant sows.