Housing systems should enable animals to interact socially while minimising aggression. This study investigated whether social interaction patterns in groups of indoor-housed gestating sows are affected by stocking density ('SD'), by varying the number of sows per pen, and day after mixing. Unacquainted sows were grouped within 5 days of insemination at 1.45 (High; 'H-SD'), 2.0 (Moderate; 'M-SD') or 2.9 m(2) per sow (Low; 'L-SD') by placing 20, 14 or 10 sows per pen. Five pens per treatment were represented, and within each pen, 10 focal sows were observed. Social behaviours between focal sows were analysed continuously for 15 min after drop-feeding the day after mixing (Day 2) and 1 week later (Day 9), and plasma cortisol and progesterone concentrations analysed on Days 2 and 26. Overall, 587 interactions were observed, with 59% being short one reciprocal actions. Sows used more physical aggression at H-SD on Day 2 such as bites (vs M-SD and L-SD: both P < 0.01), head knocks and pushes (vs L-SD: P = 0.01; vs M-SD: P = 0.06), whereas social interactions at lower SD were more frequent (L-SD vs H-SD: P = 0.004; L-SD vs M-SD: P = 0.02) and contained presumably neutral social behaviours such as nose contacts (L-SD vs H-SD: P = 0.06; M-SD vs H-SD: P = 0.07), or non-physical agonistic behaviours such as threats (L-SD vs H-SD: P = 0.07). A first-order Markov transitional analysis revealed positive feedback loops for bites and knocks or pushes at higher SD on Day 2 (all P < 0.05). Cortisol concentration decreased as SD decreased (L-SD vs H-SD: P < 0.001; L-SD vs M-SD: P = 0.02; M-SD vs H-SD: P = 0.07). Hence, interactions were more aggressive and less frequent at higher SD at Day 2 post-mixing, but differences vanished at Day 9. Valuable information can be gained from analysing the type of social behaviours and behavioural sequences, in addition to overall aggression frequency.
