Australian Journal of Agricultural Research

Publishing Australian Journal of Agricultural Research Volume 52, 2001 CSIRO 2001 A journal for the publication of original contributions towards the understanding of an agricultural system All enquiries and manuscripts should be directed to: Australian Journal of Agricultural Research CSIRO Publishing PO Box 1139 (150 Oxford St) Collingwood, Vic. 3066, Australia Telephone: +61 3 9662 7628 Fax: +61 3 9662 7611 Email: ajar@publish.csiro.au Published by CSIRO Publishing for CSIRO and the Australian Academy of Science www.publish.csiro.au/journals/ajar

Aust. J. Agric. Res., 2001, 52, 1 28 A review of the welfare issues for sows and piglets in relation to housing J. L. Barnett AC, P. H. Hemsworth AB, G. M. Cronin A, E. C. Jongman A, and G. D. Hutson B A Animal Welfare Centre, Victorian Institute of Animal Science, Agriculture Victoria, Private Bag 7, Sneydes Road, Werribee, Vic. 3030, Australia. B Animal Welfare Centre, The University of Melbourne, Victorian Institute of Animal Science, Private Bag 7, Sneydes Road, Werribee, Vic. 3030, Australia. C Corresponding author; email: john.barnett@nre.vic.gov.au Abstract. This review of sow welfare addresses all aspects of housing for adult female pigs, including the issue of piglet welfare during lactation. It puts the issue of sow welfare in perspective by briefly outlining different approaches to the scientific assessment of welfare, the feelings, preference, nature, and the functional or homeostasis approaches. We believe the last approach currently offers science the best assessment of welfare and is the approach that is utilised in this review. It involves comparing housing or husbandry systems and risks to welfare on the basis of relative changes in biological (behavioural and physiological) responses and corresponding decreases in fitness (growth rate, reproductive performance, or health/injury/immunology). The review discusses the following areas: (i) housing of individually housed pregnant pigs, with subsections on tethers, stalls, reproductive performance, exercise, and new stall designs; conventional, alternative, and outdoor group housing with subsections on aggression, electronic feeding stations, ecoshelters, and other alternative group housing designs; and other issues, such as lameness, culling, straw and other substrates, diet and hunger, quality of stockpeople, and housing around mating including oestrus detection and mating; and (ii) farrowing and lactating pigs with subsections on farrowing crates and alternative farrowing systems, stress around farrowing and during lactation, maternal behaviour and piglet survival, and sow and piglet injury and lameness. Conclusions and recommendations arising from the review include the need for public education to provide an informed consumer base that will result in some consensus on welfare issues among diverse interest groups and the need for industry education that results in better animal welfare and a sustainable industry. Some specific research recommendations include space allowance and the duration of housing for individually housed pigs, welfare issues of breeding sows in ecoshelters, piglet mortality in alternative systems, aggression in conventional and large groups, bedding, and hunger. Additional keywords: pig, piglet, sow, farrowing, fitness, welfare assessment. Introduction The welfare of farm animals generates considerable interest in Australia and internationally. Public perceptions, particularly misconceptions, of farm animal welfare issues have the potential to markedly affect the sustainability of the livestock industries. For instance, governments, industries, or consumers may react to these issues, perhaps emotively and without factual information, by introducing or changing codes of practice, banning a housing system, or boycotting an animal product. Thus, local, national, and international pressures are likely to have an increasing role in determining how animals are farmed CSIRO 2001 in the future. Furthermore, poor welfare in one livestock industry may have implications for other livestock industries as a result of generalisation of the public s views. Various aspects of pig welfare have been previously reviewed, such as stereotypies (Lawrence and Terlouw 1993), space requirements (Baxter and Schwaller 1983), and alternative housing and feeding systems (Brouns and Edwards 1992). However, other than the report by von Borell et al. (1997), there are no integrated reviews in the scientific literature and certainly none with an Australian perspective; hence this review. 10.1071/AR00057 0004-9409/01/010001

2 J. L. Barnett et al. There is a general agreement that pigs must be provided with food, water, and shelter and that a pig enterprise requires predator, vermin, and disease control procedures. Although there would be serious welfare risks if these resources/conditions were not provided, they are so basic to any pig enterprise and generally uncontroversial that they are not covered in this review. There is also general agreement on what the more controversial welfare issues are that may arise as a consequence of housing pigs. Some of the issues have been shown to result in adverse consequences and thus reduced welfare of the pig, whereas others are unresolved or may prove to be perceptions. Other welfare issues have been raised at various times and to varying degrees by some sections of the general community, scientists, and politicians. The issues most frequently raised and some reasons for raising them include the following. Confinement. Individual housing in stalls or tethers for pregnant sows and farrowing crates for lactating sows is a controversial issue because of the general lack of social contact, inability to exercise, and restricted choice of stimuli (such as other pigs and additional features of the physical environment) with which the pig can interact. Space. The space allowances provided for individually housed (confined) pigs and commonly to group-housed pigs are considered too low. Aggression. Aggression between pigs and particularly between unfamiliar pigs is a concern because of the injuries that can occur during physical contact and the stress that can arise from unresolved aggression. Stereotypies/abnormal behaviours. The presence of stereotypies is considered evidence of a poor physical and social environment and/or inadequate nutrition. Indoor/outdoor production systems. The modern intensive production piggery is considered inherently bad because of lack of space, barrenness of the environment, and the reliance on technology. In contrast, outdoor housing is considered inherently good because it provides a more natural environment and choice for the pig in performing a number of behaviours over a relatively large area, and the lower technological inputs provide for fewer equipment breakdowns that may adversely affect welfare. Thermal environment. The thermal environment poses potential risks to the pig directly in hot weather, because of the pig s inability to sweat, and indirectly in cold weather because of the increased energy requirement that may not be met. Bedding. The general lack of bedding for pregnant and lactating indoor-housed sows is an issue because it is considered to provide physical and thermal comfort and environmental enrichment by providing the pig with something to do. Bedding is generally provided in outdoor systems. Bedding may also be relevant for the stimulation of pre-farrowing nest-building behaviour. Diet. The highly concentrated diet commonly provided to pigs directly results in long periods of inactivity and abnormal behaviours and may physically damage the pigs (e.g. ulcers). The restricted diet commonly fed to pregnant pigs results in constant hunger. Foot and leg injuries/lameness. Housing indoors on concrete results in pigs becoming lame, and sharp stones and muddy conditions can cause lameness outdoors. Stockperson handling and supervisory skills. The level of skill required to manage pigs is important for all systems, but is considered more important for indoor-housed pigs because of the greater reliance on technology and the greater involvement of humans in management tasks. The greater separation of stockpeople and their pigs in outdoor systems may make it more difficult to supervise outdoor pigs. The scientific assessment of animal welfare In making a decision on whether or not an animal s welfare is seriously compromised, individuals will integrate moral views with biological facts. Thus, science has the important role of establishing the facts on how animals biologically respond to the practices under question, whether they are farming, laboratory, or general community practices for animals. However, the assessment of welfare is a controversial subject. Within scientific disciplines, variations in definitions of animal welfare exist, and, combined with variations in methodology and, in turn, interpretation, lead to disagreement (Hemsworth and Coleman 1998). It is generally accepted that there are 3 broad approaches used by scientists in studying animal welfare: the feelingsbased, the nature of the species, and the functioningbased approaches (Duncan and Fraser 1997). A more descriptive title for the third approach, the functioning-based approach, which will be used here, is the homeostasis approach. There is also a fourth approach, the animal preferences approach, which is sometimes included in the feelings approach but does not necessarily provide direct information on feelings or emotions. This approach involves studying the animal s choice for resources. We favour the homeostasis approach in assessing animal welfare. The definition that underpins this approach is The welfare of an individual is its state as regards its attempts to cope with its environment (Broom 1986a). In this definition, the state as regards attempts to cope refers to both how much has to be done by the animal in order to cope with the environment and the extent to which the animal s coping attempts are succeeding. Attempts to cope include the functioning of body repair systems, immunological defences, physiological stress response, and a variety of behavioural responses. Therefore, using such a definition, the risks to the welfare of an animal by an environmental challenge can be assessed at 2 levels: firstly, the magnitude of the behavioural and physiological responses; and secondly, the biological

Review of sow welfare research 3 cost of these responses (Barnett and Hutson 1987; Broom and Johnson 1993; Hemsworth and Coleman 1998). These behavioural and physiological responses include the stress response, whereas the biological cost includes adverse effects on the animal s ability to grow, reproduce, and remain healthy. This definition has been broadened to incorporate animal emotions (Broom 1998). We agree that animal emotions can be incorporated into the homeostatic approach as they would have evolved on the basis of their survival values and contribution to biological fitness. This concept of biological fitness generally applies to natural populations and refers to fitter animals having a greater genetic contribution to subsequent generations (Pianka 1974); this is based on their abilities to successfully survive, grow, and reproduce. Although the last attribute may not always apply to individual farm animals since reproduction is either controlled or absent for many farm animals, the ability to grow, survive, and reproduce could be considered measurements of fitness within the limits of the management system. Most production systems in agriculture have breeding and growing components and these can generate considerable data on reproductive success of individuals. For example, conception rates and mortality, morbidity, and growth of offspring can be used as a measure of fitness. Similarly, Beilharz and Zeeb (1981) and Beilharz (1982) have linked reproductive performance of domestic species with welfare. An attribute of the homeostasis approach that affords this approach credibility within scientific circles is that it contains some widely accepted criteria of poor welfare such as health, immunology, injuries, growth rate, and nitrogen balance. Furthermore, there are some excellent examples of the value of this homeostasis approach in assessing animal welfare (Hemsworth and Coleman 1998). For example, handling studies on pigs have shown that fearful pigs have a sustained elevation of plasma free corticosteroid concentrations (Hemsworth et al. 1981, 1986a; Hemsworth and Barnett 1991). The consequences of this chronic stress response in these fearful animals include depressions in growth and reproductive performance (Hemsworth et al. 1981, 1986a; Hemsworth and Barnett 1991). A counter argument is that our current knowledge may not allow detection of some of the more subtle or less serious risks to welfare. Nevertheless, less serious challenges should be reflected in biological changes, admittedly of lower magnitude, with consequent effects on fitness variables such as growth, reproduction, injury, and health. Short-term challenges can also be studied with this approach. Lay et al. (1992) studied the behavioural and physiological responses of cattle to 2 branding procedures to assess the relative aversiveness of the procedures, and Hemsworth et al. (1996) utilised behavioural and physiological responses together with growth performance to assess the welfare implications of a husbandry procedure regularly imposed (daily injections) on pigs. With our present knowledge, the homeostasis approach appears to offer science the best assessment of the welfare of animals and is the approach taken in this review. As a research tool, this approach involves comparing housing or husbandry systems, and risks to welfare are assessed on the basis of relative changes in biological (behavioural and physiological) responses and corresponding decreases in fitness. Welfare implications for housing pregnant pigs The housing of dry sows, particularly individual housing, is one of the most controversial issues of conventional pig production. Although precise figures are not known, some survey data indicate that in Australia about 62% of pregnant sows are individually housed at some stage during pregnancy, with the remainder being group-housed (Paterson et al. 1997). This is similar to the worldwide situation. In New Zealand a recent survey found that 32% of sows were housed in stalls for most of pregnancy (49% were housed in stalls at some stage during pregnancy), 40% in indoor group pens, and 28% were housed outdoors (Gregory and Devine 1999). In Europe, 70% of pregnant sows are individually housed, including 16% in tethers, and 30% are group-housed (Hendriks et al. 1998). The United Kingdom is atypical of other European countries in that no sows are individually housed and 20% are housed outdoors; in most other countries, outdoor housing is <10% or non-existent (Hendriks et al. 1998). In Sweden, because of its ban on stalls, all pigs are group-housed. There are no figures available for the USA; in our estimate approximately 60 70% of sows are housed in stalls throughout gestation. In Australia, tether housing has largely disappeared following a senate report (Anon. 1990) and a change to the Code of Practice (Anon. 1998a, 1998b) that no longer cites tethering as an approved housing method for sows. Group housing is predominantly indoors on slatted/concrete floors, in group sizes commonly ranging from 2 to around 20 sows, although it is likely that some farms use larger group sizes. Alternative group housing systems known by the authors to be in recent or current use in Australia include large indoor groups with straw and electronic feeding stations, ecoshelters (low cost sheds with bedding and large groups), and outdoor housing, although for all these systems there is little published information that relates to the Australian environment. A number of types and variations of commercial housing for pregnant pigs have been described (Svendsen and Svendsen 1997), and the welfare of intensively kept pigs has been reviewed from the European perspective (von Borell et al. 1997). Individual housing For many years there has been community concern for the welfare of pregnant pigs confined in stalls. The 2 major types of stalls are tether stalls (tethers) where pigs are housed in a partial stall and attached by the neck, or less frequently by the

4 J. L. Barnett et al. girth, to the stall by a moulded collar and chain, and cage/individual stalls (stalls) which are small rectangular pens/cages that house individual pigs. Although tethers are generally being phased out, they are still currently in use, although little used in Australia (see above). For example, in the Netherlands about 20% of pregnant pigs are housed in tethers, although they are to be phased out by 2006 (Anon. 1999). A feature frequently associated with individual housing is stereotypies; these behaviours are more common in both tether- and stall-housed sows (Broom 1983; Wiepkema 1983; Arellano et al. 1992). It has been suggested that stereotypies develop as a consequence of boredom (Fraser 1975), restraint (Cronin 1985), or a frustration of feeding motivation (Lawrence and Terlouw 1993). These last authors have reviewed the literature, particularly on the contribution of feeding motivation to stereotypies. There has been and continues to be considerable controversy on the causation and function of stereotypies in farm animals. Excessive chain manipulation by sows is a stereotypy seen in gestating sows housed on tethers, and Lawrence and Terlouw (1993) have shown that food restriction contributes to the development of this stereotypy. Unavoidable fear or stress and barren and restrictive environments have also been implicated in the development of other stereotypies. Mason (1991) refers to examples of bodyrocking in mentally handicapped patients when distressed and where the incidence of stereotypies increases with increasing confinement. Cooper and Nicol (1991) have proposed that some forms of stereotypies reduce responses to aversion by affecting the animal s perception of the situation. Thus it is clear that different forms of stereotypies may have different causes, such as frustration, stress, and lack of control and stimulation; however, our understanding of the motivational basis of stereotypies is poor. A similar controversy exists in relation to the function of stereotypies. Based on early evidence of associations between stereotypies and physiological signs of coping such as reduced corticosteroid concentrations, reduced adrenal gland weights, and reduced ulceration, there is a view that stereotypies may be a coping response. However, more recent studies and re-interpretation of some of the early evidence question this general coping hypothesis for at least some forms of stereotypic behaviours (Mason 1991; Rushen 1993). Furthermore, although some evidence exists to indicate that stereotypies may be coping mechanisms in the short term, it is unknown whether they exert benefits in the long term. Irrespective of the function of stereotypies, the existence of a stereotypy is indicative at the least of a past problem for the animal in coping with its conditions. Stereotypies that result in physical damage to, or illness in, the animal (e.g. the development of lesions in stall-housed sows that persistently rub their tail roots from side to side against stall fittings or wind-sucking in horses where persistent wind-sucking can lead to gastrointestinal catarrh and colic) have obvious and immediate implications for the welfare of farm animals. Thus, although stereotypies should not be used alone, they can be used together with other biological responses and their consequent effects on biological fitness, to assess risks to animal welfare. Community concern about confinement, the presence of stereotypies, the evidence of a chronic stress response in pregnant pigs housed in tethers (Barnett et al. 1985, 1987a, 1987b), and the general lack of use of this housing system in Australia all probably influenced the recommendation that tethers should be banned and that alternatives to individual confinement of pregnant pigs are preferred in Australia (Anon. 1990). In the European Union, tethers are to be phased out by 2006 (Directive 91/630/EEC; von Borell 1996). Sweden has already banned stalls, and new regulations in Norway defining space allowances for sows and thereby effectively prohibiting stalls come into effect in 2000 (Burke 1996). Tethers Notwithstanding the very limited use, if any, of tethers in Australia, it is worthwhile briefly reviewing the research on tethers as it both validates the approach outlined in the section on welfare assessment and provides some insights into the complexities of making recommendations on welfare. A number of studies have shown evidence of a chronic stress response of pigs housed in tethers (Barnett et al. 1984, 1985, 1987a, 1987b, 1989, 1991). Furthermore, some studies showed consequential effects indicative of reduced welfare, such as a change in nitrogen metabolism indicative of gluconeogenesis and a metabolic cost (Barnett et al. 1985), an increased metabolic rate (Cronin et al. 1986b), immunosuppression (Barnett et al. 1987b), and a reduced reproductive performance (Barnett and Hemsworth 1991). Some of these data were probably used as the scientific basis for the withdrawal of approval for tethers in Australia (Anon. 1990), although many factors other than science were involved in that decision (Barnett 1991). Notwithstanding the finding that tethers can adversely affect welfare, this does not mean that tethers inherently result in poor welfare. Subsequent research showed that the design of the tether stall divisions had a significant impact on the magnitude of the chronic stress response (Barnett et al. 1987a, 1989). Thus, tether stall divisions in which vertical bars separated neighbouring pigs resulted in a chronic stress response, whereas, if the bars were covered in steel mesh, the free cortisol concentrations were similar to those in grouphoused pigs. It was concluded that a poor design of tether stall division (i.e. vertical bars) resulted in unresolved aggressive behaviour, demonstrated by an increase in retaliatory responses to an aggressive interaction (Barnett et al. 1987a, 1987b). Changing the design of the tether stall divisions such that there was still visual and some limited contact between neighbouring pigs abolished the behaviourally

Review of sow welfare research 5 induced chronic stress response and ameliorated the risk to welfare from the housing system per se (Barnett et al. 1987a, 1989). Other studies (Becker et al. 1985; Friend et al. 1988) have shown acute responses to tethering, although the latter study found no long-term adverse effects of different housing systems. The design of the tether stall divisions in these two studies was not described. Janssens et al. (1994) showed increased responsiveness to adrenocorticotropic hormone (ACTH), indicative of a chronic stress response (see Hennessey et al. 1988 and Rushen 1991 for rationale and review of ACTH responsiveness), in tethered gilts, particularly when visual and tactile contact were reduced with a solid division, compared with group-housed pigs; however, this finding needs further evaluation as Barnett et al. (1981) have shown that visual isolation of individually housed gilts in a pen also results in a chronic stress response. Stalls A common housing system for pregnant pigs is stalls, which, like tethers, were introduced predominantly to control feed intake and reduce aggression. Although precise values of the percentage of pregnant pigs housed in stalls and the time spent in stalls are unknown, limited survey information suggests that 26% of sows are stall-housed in Australia for most of their reproductive cycles (i.e. except for farrowing in crates and a period of group housing around mating) and up to 62% may be in stalls for part of their reproductive cycle (i.e. in stalls for a restricted time followed by group housing, farrowing in crates, and group housing around mating) (Paterson et al. 1997). The majority of sows would also be housed in groups for about 1 week after weaning, for the purpose of being remated. In The Netherlands about 75% of pregnant pigs are housed in stalls. Community concern still exists for the practice of housing pregnant pigs in individual stalls (Anon. 1992; Baker 1996). In contrast to the findings with tethers, early work with stalls showed that there was no physiological evidence that stalls (of certain designs) were associated with a risk to the welfare of pregnant pigs. In 2 experiments, Barnett et al. (1989) found that pigs housed in stalls, with either vertical bars or wire mesh on the front section of stall divisions, showed a moderate but significant increase in basal free cortisol concentrations compared with pigs housed in groups, but this increase was markedly less than that of pigs housed in tethers. Furthermore, whereas glucose concentrations were elevated, indicative of a metabolic cost, in pigs housed in tethers, no increase was evident in pigs in the 2 stall treatments in either experiment. Thus, although some of these data provide prima facie evidence of a stress response in stalls, there was no apparent adverse consequence and thus no significant risk to welfare. In a study over a number of parities, Broom et al. (1995) concluded that by the fourth parity, welfare was reduced in stalls compared with grouphoused sows, on the basis of increased stereotypies, increased aggression, and reduced body weight. However, they were unable to find any differences on the basis of physiological, immunological, or reproductive measures. Similarly, von Borell et al. (1992) found no differences in ACTH responsiveness between sows in stalls and groups. A comparison of differential lymphocyte counts from sows housed in stalls and groups during gestation showed no treatment effects at farrowing (Nind et al. 1997a), suggesting that the immune systems were not differentially activated. Sows or gilts in stalls are less responsive to external stimuli, including water poured on their back, sow grunts, piglet squeals, and an electronic buzzer, than group-housed pigs (Broom 1986b; Barnett 1995), although the reasons for this reduced responsiveness are unknown. Nevertheless, there was evidence that stall design may adversely affect the welfare of pregnant pigs (Barnett et al. 1991). Pigs in stalls with only horizontal bars on the stall divisions showed evidence of a chronic stress response, based on a sustained elevation of basal free cortisol concentrations, similar to that seen in tethered pigs, and active avoidance of neighbouring pigs. Pigs housed in stalls comprising vertical bars showed cortisol concentrations similar to group-housed pigs (and lower than pigs in both tethers and stalls with horizontal bars). Surprisingly, these latter stallhoused pigs had high levels of aggressive interactions with their neighbours. The above physiological data on both tethers and stalls have challenged some conventional thoughts on individual housing and one reasonable conclusion is that it is the design of the housing system that is important to welfare rather than the housing system per se. A recent modification to the Code of Practice for pigs (Anon. 1998a) excludes feeder and drinker facilities from the length of the minimum stall size (2.0 by 0.6 m). Although there are no data on stall dimensions in Australia, a recent small survey of stalls in New Zealand shows that average stall sizes are 1.91 m long and 0.58 m wide (range = 1.63 2.50 m for length and 0.45 0.69 m for width; Gregory and Devine 1999). Thus, it is likely that some stalls in Australia do not comply with the minimum requirements of the Code of Practice. However, the implications of stall dimensions for welfare are not known. The dimensions of stalls are largely based on both the size of pigs (Baxter and Schwaller 1983; Curtis et al. 1989a) and the requirement that they not be able to turn around (Robertson et al. 1972). Individual housing and reproductive performance Although housing in tethers adversely affects reproductive performance (Barnett and Hemsworth 1991) compared with group housing, the data on individual v. group housing are equivocal, on the basis of weaning to mating interval and mating, conception, or pregnancy rates. A number of studies have reported that sows housed in groups have a shorter weaning to mating interval than tether- or stall-housed sows

6 J. L. Barnett et al. (Hemsworth 1982; Schmidt et al. 1985), although some authors found contradictory results (Karlberg 1980; Lynch et al. 1984; Vessuer et al. 1994). Barnett and Hemsworth (1991) found that in 15 studies reviewed, 8 showed better reproduction in group-housed pigs, whereas only 4 showed better reproduction with individual housing. The behavioural disturbances and the stress response often associated with tethers (Jensen et al. 1970; Barnett et al. 1985) may be responsible for the delay in the onset of oestrus in the sow in individual housing systems. However, it should be noted that in groups, sows can be exposed to stress resulting from low social rank and aggressive interactions (Vessuer et al. 1994). In a recent study in an Australian piggery, a comparison was made of about 220 sows housed in stalls for 5 weeks post-mating prior to being housed in groups and about 450 sows housed in groups; measurements were taken at 4 different times. Notwithstanding the constraints of the comparison involving different units within the farm and the limited statistical power (degrees of freedom = 6), the stall-housed sows had more piglets born alive (11.6 v. 10.8; P < 0.05; Agribiz Engineering 1999). This study suggests that there may be some advantage in terms of biological fitness, and hence welfare, from stall housing, at least for a limited time. The survey data of Paterson et al. (1997) similarly showed improved overall performance, on the basis of a lower removal rate due to a combination of reproductive failure, lameness and locomotor problems, age, death, and euthanasia, in farms that had both pens and stalls compared with those that only had pens. Our interpretation of pens and stalls in the study of Paterson et al. is that after mating, sows were housed for a time in stalls followed by the remainder of gestation in groups, although they did not define pens and stalls. The consequences of housing in stalls for shorter or longer periods of time and the advantages/disadvantages for overall welfare remain to be determined. Because stall housing is a controversial issue from the view of public perception, housing in stalls for a defined period that is considerably less than the period of gestation may be a reasonable compromise. Exercise and new designs of stall Exercise during gestation is associated with a reduced farrowing time and improved piglet survival to weaning (Ferket and Hacker 1985) and a reduced incidence of lameness in gilts (Hale et al. 1984), and it would intuitively be expected that individually housed pigs have less exercise than grouphoused pigs. Indeed, muscle mass and bone strength are reduced in pigs housed in cage stalls over successive pregnancies (Marchant and Broom 1996a) and joint damage was increased in individually housed compared with grouphoused pigs (Fredeen and Sather 1978). Similarly, higher resting heart rates (Marchant et al. 1997) and the longer time that sows take to lie down (Marchant and Broom 1996b) in stalls compared with group-housing have been interpreted as due to lack of physical fitness (Marchant et al. 1997). The longer time to lie down in stalls is also considered a consequence of the spatial requirements of the sow (Baxter and Schwaller 1983). However, the relationship between exercise and reproductive performance is unclear and one reason may be that pigs do not appear willing to exercise voluntarily (Blackshaw and McVeigh 1986). Nevertheless, a common criticism of individual housing is that such confinement prevents the pig from freely moving around and by implication there is concern over the lack of ability to exercise. Some recent design innovations have resulted in a stall that allows pigs to turn around (McFarlane et al. 1988; Johnson et al. 1990). One commercial type of turn-around-stall, the Moorcomfort gestation stall, has side panels that are hinged at about 60 cm from the front, creating a swing partition between the rear two-thirds of adjacent stalls, allowing neighbouring pigs to borrow space from each other and thus turnaround. A small study showed that cortisol concentrations were similar to those of group-housed pigs (Barnett and Taylor 1995). Group housing conventional Indoor group housing is a common housing system for pregnant pigs, and whereas some attention has been given to factors such as space allowance and group size (Jensen et al. 1970; Ford and Teague 1978; Kuhlers et al. 1985; Barnett et al. 1986; Hemsworth et al. 1986b), particularly for reproductive performance, less consideration has been given to other factors such as social contact, dominance order, and design features in pens that may affect welfare. A common criticism of individual housing systems for pigs is that social contact is disrupted. However, the effects of social rank on reproductive success of group-housed sows indicate potential problems for certain animals. For example, Mendl et al. (1992) reported that socially intermediate pigs had higher concentrations of salivary cortisol, were more responsive to an ACTH challenge, indicative of a chronic stress response, and had lighter piglets. Social rank during pregnancy can also affect maternal behaviour, with subordinate sows subsequently displaying more stereotypies, increased restlessness, and more interrupted suckling bouts than dominant sows after farrowing (Csermely and Nicosia 1991). Similarly, Nicholson et al. (1993) reported that, compared with dominant and submissive sows in the same group, socially intermediate sows showed specific signs of stress (elevated cortisol and reduced natural T killer-cell activity) and had lower farrowing rate and smaller litter size. Other factors such as space allowance are also likely to be involved. Recommendations for space requirements for adult pigs are few, probably based on current practice, and are in the range of 1.4 1.8 m 2 /pig (Cale 1979; Anon. 1998a, 1998b). There is clear evidence of a chronic stress response and reduced reproductive performance if space allowance is insufficient (e.g. 1 m 2 /pig; Hemsworth et al. 1986b;

Review of sow welfare research 7 <1 m 2 /pig; Barnett et al. 1992). Although the former study indicated that there may be reproductive performance advantages of housing at 3 m 2 /pig over 2 m 2 /pig, the physiological criteria indicated no differences between these space allocations. None of the recommendations takes into account the amount of additional free space available to pigs kept in large groups and the potential to reduce space allocation per pig in such group pens, and this aspect warrants research. Some limited research by Taylor et al. (1997) has shown that varying group sizes, of 5, 10, 20, and 40 sows with a space allowance of 2.0 m 2 /sow, had no effects on reproductive performance (proportion of sows that farrowed, piglets born per sow, and piglets born per sow alive, stillborn, or mummified). Although aggression, which was measured on Days 1 and 2 after grouping, increased as group size increased, the number of lesions, measured on Days 5 and 53, were similar across treatments. In the same study, reducing space allowance for groups of 10 sows from 2.0 to 1.2 m 2 /sow increased aggression. Similarly, Olsson et al. (1994) reported increased injuries as group size increases and Weng et al. (1998) reported increased aggression and injuries with decreasing space allowance. The latter study recommended a space allowance between 2.4 and 3.6 m 2 /sow for groups of 6 pregnant sows. The latter study also emphasised that the results could not be extrapolated to other group sizes and space allowances. There are no recommendations on group size for adult pigs in the Codes of Practice relating to welfare (Anon. 1998a, 1998b). Nevertheless, this management factor may vary widely in commercial practice and may affect both welfare and sexual behaviour. Studies by Barnett et al. (1984, 1986) showed that housing of sexually mature gilts in pairs resulted in a chronic stress response compared with housing in groups of 4 8. Both large group size (24 v. 8 pigs) and small group size (3 v. 9, 17, or 27 pigs) may have detrimental effects on oestrus expression (Christenson and Ford 1979; Christenson and Hruska 1984), and increasing group size and concomitantly decreasing space allowance may have detrimental effects on oestrus expression (Cronin et al. 1983). Broom et al. (1995) compared sows in groups of 5 fed in stalls and a group of 38 sows that had an electronic feeding station. Although there was increased aggression in the larger group, particularly after initial mixing, any differences in aggression and stereotypies had disappeared by the fourth parity. Further research is required to determine the optimum group size for pregnant pigs. There are no data on space allowance/group size interactions for adult pigs. Aggression Notwithstanding the general lack of research at the time, the Senate Enquiry (Anon. 1990) recommended that individual stalls should be installed into group pens on the basis that they satisfied both the farmer s requirement to individually feed pigs and the social requirements of pigs. This design also appeared to have the added advantage of providing escape areas for subordinate pigs during bouts of aggression. Petherick et al. (1987) have shown some advantages of partial stalls in reducing aggression around feeding in pens of grouphoused pigs. However, whether the reduced aggression was due to individual feeding or the provision of escape areas was not determined. Barnett et al. (1992) subsequently demonstrated welfare advantages of partial stalls within group pens based on a reduction in aggression immediately after grouping unfamiliar pigs and in the longer term by feeding in the partial stalls. Feeding in the partial stalls also resulted in longterm benefits, based on lower free cortisol concentrations and a higher cell-mediated immunity. However, a subsequent study examining effects of pen design in reducing aggression in groups showed no benefits of partial stalls within pens either in the short term or subsequently (Barnett et al. 1993a), although in that study the pigs were not fed in the stalls. Thus, it would appear that partial stalls may confer advantages in reducing aggression if the pigs are routinely fed in the partial stalls. Similar design modifications, i.e. incorporating stalls into group pens, have been reported elsewhere (Svendsen and Bengtsson 1983; Edwards 1985). Pen shape has some effects on aggression (Barnett et al. 1993a). Aggression was less in rectangular pens than square pens as a result of grouping unfamiliar pigs, provided the space allowance was 1.4 m 2 /pig; the benefits were lost in larger pens providing 3.4 m 2 /pig. However, Olsson et al. (1994) recommended against using long narrow pens with liquid feeding on the basis of competition and variable intakes. Aggression among recently grouped unfamiliar gilts and sows is seen as a welfare disadvantage of group housing. Aggression can be reduced in gilts by: (i) modifying pen size and shape (Barnett et al. 1993a) on the basis that pigs require a minimum space in which to fight; (ii) modifying pen design (Petherick 1985; Petherick et al. 1987; Barnett 1997) on the basis that the provision of escape areas reduces aggression; (iii) pre-exposing pigs to auditory and olfactory stimulation in their new pen (Kennedy and Broom 1996); (iv) grouping after dark (Barnett et al. 1994), on the basis that it is the normal sleeping time, or providing feed ad libitum (Petherick 1985; Barnett et al. 1994) on the basis that restrictively fed pigs may prefer to feed than fight; (v) using masking odours (McGlone et al. 1981; McGlone 1985; Leuscher et al. 1990) on the basis that anosmic pigs show reduced aggression (Meese and Baldwin 1975); and (vi) using mood-altering drugs (Barnett et al. 1993b, 1996) on the basis of their positive effects in animal models (Gustafsson and Christensson 1990). However, all or some of these methods may only be effective in postponing aggression rather than reducing it. There are few rigorous recommendations and this subject needs further research. Notwithstanding the possible welfare advantages of reduced aggression from pen designs for group-housed sows that incorporate partial feeding stalls within the group pen,

8 J. L. Barnett et al. the evidence that reduced space allowance can compromise welfare (Hemsworth et al. 1986b) suggests a potential compromise between space and pen design. An experiment comparing feeding in full or partial stalls and feeding troughs and space allowances of 1, 1.4, and 2 m 2 /pig that were additional to the area occupied by the stalls/troughs showed that full stalls reduced the level of stress and aggression around feeding and that 1 m 2 /pig was inadequate on the basis of elevated cortisol concentrations and reduced immunological responses (Barnett 1997). The author recommended a minimum space allowance of 1.4 m 2 /pig in addition to the provision of full stalls for gilts weighing 117 kg. If space was a limiting factor in such situations, partial stalls appear to confer some welfare benefits. It is not known if more space is required for larger sows. Alternative indoor housing Modifications to indoor pens In addition to the modifications outlined above, there are other modifications that have been applied to conventional pens. However, some have been pilot trials with little rigorous scientific evaluation, whereas others have been more scientifically tested and either they have not been reported or perhaps they have not been adopted. For example, a mazesystem providing additional space from large pens with a number of partial partitions reportedly results in less fighting and associated injuries (Nehring 1981). The concept of trickle-feeding to reduce aggression is still being used (Wit 1996), and Barnett and Taylor (1997) have shown the importance, based on acute and chronic stress responses, of providing feed to pigs either concurrently or in a systematic manner. Modifications of the stall/group system are covering the stalls to provide kennels or providing a separate enclosed (group) cubicle in addition to the stalls (Edwards 1985). Scheepens (1990) has designed a specific stress free system that does not require pigs to be regrouped, provides additional space, and uses liquid feeding supplemented with substrates such as chopped straw for the non-feeding period, although there have been no reports following its introduction. Similarly, the family pen housing system maintains pigs in family groups with areas for rooting and other activities, straw, space, a nest, individual feeding places with straw for the non-feeding period, and escape areas (Stolba and Wood-Gush 1984; Wechsler 1996). A number of alternative group housing systems have been described (Svendsen and Svendsen 1997). Electronic feeding stations and other non-conventional housing systems An alternative to small groups of sows in conventional pens is to use large groups and these are frequently associated with electronic feeding stations. These systems usually involve about 40 sows per feeding station housed in either stable (Edwards 1985; Edwards et al. 1988) or less stable groups whereby pigs due to farrow are regularly removed and replaced by pregnant pigs (dynamic groups; van Putten 1990; van Putten and Burgwal 1990), varying from 40 to 300 sows. The system has the advantage of controlling feed allocation to individual sows. Electronic feeding stations have been in use for about 15 years in the pig industry and some of the early problems of vulva biting (van Putten and Burgwal 1990), due to aggression around the feeder, have been solved by changing the design of the feeding station, although aggression can still be a problem (Edwards et al. 1988). Although feeding order appears stable in fixed groups of sows (Hunter et al. 1988), it is less stable in dynamic group housing systems (Bressers et al. 1993), thus making it difficult to identify animals with a health problem on the basis of their feeding order. Also, if feeding order is not stable, this may result in a chronic stress response (Barnett and Taylor 1997), although this has not been determined in systems using electronic feeders. There are also reports of variation between herds and some loss of reproductive performance (Jennings 1988; Bokma 1990) and high levels of aggression (Hunter and Smith 1991) in these systems. If the use of electronic feeding stations results in increased variation between herds, it suggests that a high level of training and management is required for their successful use. As with any technology, there is a reliance on systems operating correctly; there are reports of lost electronic identification collars (Edwards 1985) and ear trauma from chewing on pigs with electronic ear tags (Sherwin 1990). Implantable transponders should overcome these problems (Merks and Lambooy 1990; Lambooij et al. 1995; Stark et al. 1998). There are limited reports of the use of electronic feeding stations for pregnant pigs in Australia (Taylor and Clarke 1988). Another housing system involving electronic identification is one involving continuous housing in small groups with controlled access to a number of feeding compartments and some bedding within the main area (Morris and Hurnik 1990). A comparative study of gilts in this system and conventional group pens showed no significant reproductive benefits (e.g. total born or weaned) or physiological changes (e.g. cortisol concentrations or immunological measures) associated with the design changes (von Borell et al. 1992). Ecoshelters A recent innovation for housing pregnant pigs has been the use of ecoshelters or hoop structures. They have been used in the past to house large groups (several hundred) of grower/finisher pigs and more recently there has been interest in using them for all stages of production. They are considered low cost buildings which rely on bedding to absorb faeces and urine as well as for use when sleeping. They generally have a concrete platform for feeders and drinkers and a compacted surface for the deep bedding. They can house groups varying from <20 to several hundred sows. They are frequently open-ended buildings with gates to prevent the

Review of sow welfare research 9 pigs moving out of the sheds and have plastic hoop roofs. There are no data available to compare performance between systems, although there are some reproduction data on ecoshelter systems. Small units have been developed in Australia, Canada, and the USA for experimental purposes, and in Australia there are some commercial units being trialed. In overseas studies, reproductive performance was comparable with conventionally housed pigs (Connor et al. 1997), whereas in an Australian study there was an unacceptably high preweaning mortality (Payne 1999). The potential of ecoshelters for breeding pigs warrants a thorough examination because of their possible welfare and production implications. Group housing outdoors Although large-scale commercial pig production outdoors is not a new method of housing, over recent years there has been a renewed interest in outdoor pig production systems (Edwards and Zanella 1996). In the United Kingdom and other countries in Europe, economic circumstances in particular as well as concerns about animal welfare in intensive indoor systems have led to a rapid increase in outdoor production (Stark et al. 1989). At present, sows living in outdoor conditions account for 20 25% of breeding sows in the UK, 2 4% in Denmark, and 1% in both Italy and the Netherlands (von Borell et al. 1997; Hendriks et al. 1998), and for 7 9% of sows in France (Berger et al. 1998; Hendriks et al. 1998). Current estimates in Australia are 5 6% (I. Farran, pers. comm.) with a projected estimate of 7 10% by the year 2005 (Agribiz Engineering 1999). In New Zealand, 28% of sows were housed outdoors (Gregory and Devine 1999). The welfare issues in outdoor pig production have been reviewed by the UK Farm Animal Welfare Council (Anon. 1996), although that report generally appears to be based more on industry practice than rigorous science. The report details various concerns including health and disease, access to food and water, stocking density, paddock rotation, mutilations, choice of breed, predation from foxes and birds, and the use of electric fences. The report has 48 recommendations either for setting allowed practices or for the need for more research, and forms the basis of this brief review. A number of welfare concerns identified by the Farm Animal Welfare Council (UK) are unique to the outdoor system, such as behavioural startle due to over-flying hot air balloons and which may result in injuries and death (Penny et al. 1995). There are no reports of this problem, to date, in Australia. Similarly, although an occasional lowflying aeroplane or helicopter could be expected to startle the pigs, it is likely that they would become habituated if it was a regular occurrence, unless such occurrences were highly aversive. It is common practice to reduce the amount of foraging behaviour shown by outdoor pigs by placing metal rings in their noses, which supposedly cause discomfort when they root. This procedure may become undesirable on animal welfare grounds. As well as inhibiting rooting behaviours, nasal ringing affects grazing, stone chewing, and other behaviours that may have welfare implications (Horrell et al. 1997). Horrell et al. (1997) suggested that the motivation to root may be independent of hunger and that providing other opportunities to root ameliorates the reduction in welfare due to ringing. However, pasture management is also an important consideration associated with outdoor pig production. A strategy to reduce pasture damage in sows without nose-rings by providing increased roughage in the form of sugarbeet pulp was unsuccessful (Braund et al. 1998). Although it has been implied that the adverse welfare effects of nose-ringing may be offset by the fact that this management practice makes the keeping of sows outdoors more economically viable (von Borell et al. 1997), further work is required on the welfare implications of nose-ringing. Few specific studies have been conducted on nutrient requirements of outdoor pigs (McGlone 1997). From the point of view of pig welfare, the most important requirement is to ensure that sufficient food reaches all animals and this is generally achieved by widely spreading the feed on the ground. The increased social stress seen among indoor sows at feeding is not observed among outdoor sows (Martin and Edwards 1994; McGlone 1997). McGlone (1997) has argued that the quickest way to get the sow into the wallow (and thus prevent her core temperature from rising) was to not insulate the hut, although insulated huts are important for the survival of piglets (McGlone 1997). However, if lactating sows spend too long away from the litter this could compromise their welfare, as water is not located within farrowing huts. Although hut design for winter, in colder climates, would probably favour insulation of the hut to conserve the sow s body heat, little research has been done to determine the space requirements in the huts for breeding and gestating sows. Tober (1996) suggested that multiparous sows required 1.3 m 2 per sow, whereas primiparous sows needed only 0.95 m 2 per sow. For dry sows, Edwards and Zanella (1996) reported that a housing space allowance of 1.2 1.5 m 2 per sow is recommended to ensure that the more timid sows gain access. The design, number, and size of huts/shelter for dry sows require further research. Unsuitable sites and climate, particularly exposed wet and windy sites, greatly increase the potential for poor welfare in outdoor systems (Thornton 1990; Anon. 1996). Recommendations for the UK are that paddocks should be on relatively flat land in a low-rainfall area on a light topsoil overlying a free-draining subsoil with the absence of sharp stones likely to cause foot damage (Thornton 1990). Recommendations in Australia are that outdoor production should be confined to areas that experience few days over 30 C and even fewer which exceed 35 C, a low rainfall (although 38 63.5 cm is workable provided there is adequate drainage), gently sloping land to reduce the risks of flooding