Changing legislation and consumer attitudes towards the welfare of farm animals means that individual housing of dry sows will soon be illegal in the United Kingdom. Other European countries, and in particular those for whom the UK is an important export market (such as The Netherlands), are under pressure to follow suit. Group housing is generally believed to provide a higher degree of welfare than individual housing, because it allows the animal greater behavioural freedom. However, group housing does not eliminate behaviours such as stereotypies, which are associated with an inadequate environment and have long been the cornerstone of the (ethological) argument against individual dry sow housing. Also, group housing promotes the occurrence of agonistic behaviour, potentially reducing the welfare of most, if not all animals in the group. The present thesis addresses these undesirable behaviours in group housed sows, and investigates the role food level and the provision of straw may have to play in trying to reduce them. It also looks into the possibility of developing tests to identify gilts which are likely to develop these behaviours during their adult life, by assessing individual behavioural responses to a number of challenges before the animals reach puberty.In Chapter 1 the hypothesis was tested that food motivated pregnant sows need a substrate, for example straw, to express foraging behaviour. The absence of a suitable substrate may under certain conditions result in the development of abnormal oral activities, such as stereotypic chain and bar manipulation. The study described used 96 gilts, all between 1 and 3 weeks post service, which were entered into a 2 × 2 factorial design comparing food level (Low 1.8 kg/23 MJ day -1and High 3.2 kg/40 MJ day -1) and the provision of a foraging substrate (Straw and No straw). The gilts were loose housed in groups of 6 with individual stalls. Behaviour was recorded over the first two parities, by time sampling for the 2 and 6 hours after the start of feeding and over 24 hours using video recordings.The results show that activity levels were highest just after feeding, with low fed sows being more active than high fed sows. Most of the postprandial activity consisted of manipulating substrates. In low fed sows without straw, particularly in parity 2, this behaviour was mainly directed towards chains and bars, resulting in levels 3 to 4 times higher than in other groups. Low fed sows with straw directed their foraging behaviour mainly towards the straw. Chapter 1 concludes that in food restricted pregnant sows abnormal high levels of chain and bar manipulation can be prevented by providing straw which apparently acts as a foraging substrate.The second chapter described how the treatments applied in Chapter 1 affected the performance of the animals as well as their chronic and acute physiological stress responses. The latter was tested by injecting the pigs at the end of their second pregnancy with adrenocorticotrophic hormone, ACTH (1 iu kg -1live weight), which induces the adrenal cortex to release all available cortisol. Saliva samples were taken before and after the ACTH challenge, and baseline and response levels of salivary cortisol were assessed.The results emphasise the importance of straw: significantly more non-straw sows failed to start the second parity than straw sows. Food level showed expected differences in weight gain, with high fed sows gaining more weight and back fat over both pregnancies than low fed sows. Low fed animals with straw tended to gain more weight and back fat over both pregnancies than low fed sows without straw. Litter performance was not influenced by the straw treatment. High fed sows without straw were found to have the highest concentration of basal cortisol, but no treatment effects were found in increase of salivary cortisol concentrations in response to the challenge with ACTH. The chapter concludes that provision of straw may buffer the adverse effects of a low food level on weight and back fat gain in group-housed pregnant sows, but that it has no effect on reproductive performance. The results of the ACTH challenge tests suggest that physiological differences between treatments are most likely to be of an acute nature, rather than a chronic one.In Chapter 3, the focus of attention is moved towards large groups of dynamic sows. It addresses the hypothesis that hunger may be a cause of increased aggression between animals. For this study, sixteen groups of 5 gilts were introduced over 8 months to 1 of 2 dynamic groups in a deep straw yard. Sows in each group were receiving a different level of food from an electronic sow feeding system: High (3.0 kg / 38 MJ day -1) and Low (1.6 kg / 20 MJ day -1). The behaviour of the animals was recorded throughout both pregnancies using a time sampling technique. The total number of animals in each dynamic group was maintained at around 30, through the use of additional sows. The results indicate, as expected, that food level affects body weight: high fed sows were heavier than low fed sows.However, no effect of the food treatment was found on litter size or litter performance. Behaviour observations showed a low fed sows to be more active and manipulating the straw more than H sows. However, food level did not affect the number of aggressive interactions, or the level of skin damage on the sows. In contrast, the introduction of new animals to the groups did: pigs were more involved in aggressive interactions on the day they were introduced, than on other pig's introduction days and no introduction days. Average skin lesions per experimental sow were also higher in the days immediately following introduction. The conclusion of Chapter 3 is that in a sequential feeding system with plentiful straw, aggression is not influenced by the level of food. In these systems, the major factor giving rise to aggression is the introduction of new animals to the resident group.In Chapter 4 the spatial organisation of the dynamic groups of pigs used in the previous chapter was investigated more closely. Although Chapter 3 suggested that food level does not have an overriding effect on aggression, an increased food motivation may well affect the use of the pen by the animals, through changes in personal space requirements and / or the "defense" of particular areas. In order to do this, the spatial organisation of the animals was recorded throughout both pregnancies on a floor plan of the building, at fortnightly intervals. From these a range of spatial parameters were calculated: the frequency of visiting predetermined pen areas, the average location of the pig in the pen, the average inter individual distance, the distance and identity of the nearest neighbour and the proportion of observations an animal spends within a two meter range of another animal.The effects of food level were restricted to an increased use of the drinker area by high fed sows, and an increased use of the feeder area by low fed sows. The time since introduction of a new subgroup affected all spatial parameters, with the average inter-individual distance between resident and new animals decreasing, and that within subgroups increasing. The chapter concludes that social integration appears to be a staged progress, starting off with a peak in agonistic behaviour, followed by an increased use of the same areas of the pen. The final stage, which involves a random sharing of the immediate area around an individual, was not achieved before pigs were taken out of the group for farrowing.The fifth chapter described an investigation into the consistency of a gilt's behavioural responses between and within test situations, aiming to relate them to the development of undesirable behaviours in the group housed adult animal. To achieve this, behavioural responses of gilts were tested under 4 different circumstances. Gilts were housed in two series of 16 groups each (n= 6 or 7 gilts per group). The test situations were: Situation 1: an open field with a novel stimulus (bucket or human); Situation 2: individual access to food for 15 minutes after a 20 hour period of food deprivation; Situation 3: competition for food after food deprivation; Situation 4: general activity and feeding behaviour in a group over a 24 hour period.Situations 1 and 2 were assessed 4 and 3 times over a 2 and 1 week period, respectively. Both showed high levels of consistency in the behavioural responses of the gilts. Principal components analyses was used to reduce the number of variables per test situation and facilitate measurement of consistency across test situations. The amount of variation explained by the first component was generally more than twice that explained by any subsequent components. The only significant correlation between factor scores calculated from the first components was found between Situation 1 and 2 in the first series. None of the calculated factor scores showed bi- or multimodal distributions. The chapter concludes that, whilst over a short period of time gilts respond consistently to a specific challenge, they do not display the same consistency when challenged in a different context. This lack of inter-situation correlations, plus the absence of multimodal distributions, fails to support the view that behavioural "types" of gilts exist.The General Discussion addresses the practical implications of the findings in Chapters 1 to 5 in three parts. Part 1 suggests that group housing may result in reduced development of stereotypic behaviour, even though factors such as food level and straw provision may have an even greater effect. It also queries the relationship between the feeder area and the development of abnormal repetitive behaviour: does the association with food reinforce of this area reinforce the behaviour? Finally, on the subject of stereotypies, it highlights the need for continued research into alternative substrates which can be used in circumstances were straw is not an option. In Part 2, the discussion on large dynamic groups starts off by highlighting the fear of many producers who have to convert to group housing, that aggressive behaviour will dramatically reduce the welfare of the animals. It suggests that dynamic group housing systems, in which aggression is an inevitable consequence of the constant restructuring of the social organisation, should preferably not be operated in the absence of a substrate.Further more, the frequency of introducing sows to an existing group should be kept to a minimum, for example by operating semi-dynamic groups. In the final section of Part 2, the stability of the subgroup is discussed, and the concept of incorporating separate lying areas supported. Part 3 reaffirms that pigs are individuals, and that their individuality might be based on differences in a range of underlying character traits. It questions the feasibility of identifying gilts unable to "cope" with challenges in adult life, although it may well be possible to identify behavioural or physiological predictors of undesirable behaviour shortly before they start developing. Finally, it acknowledges the fact that individual behavioural differences between pigs are both unavoidable as well as desirable, but that they do demand a higher level of stockmanship if sows are kept in groups, compared to individually housed animals.