The experiment was conducted on a breeder unit with experimental sows (Camborough 42, PIC Australia, Grong Grong, NSW, Australia) mated in July and September 2021 and farrowed over two replicates in November (maximum average daily temperature 25.0 ± 5.9°C; minimum average daily temperature 10.7 °C± 4.4°C) and January (maximum average daily temperature 30.8 ± 4.7°C; minimum average daily temperature 16.5°C ± 4.6°C). One hundred and seventy-one sows (parity 2.6 ± 0.12) were observed from entry to the farrowing house until weaning. Sows were moved into the farrowing house 4.8 ± 0.21 days prior to farrowing. Farrowing sheds were cross-ventilated, with a temperature (>28°C) activated dripper system above every farrowing space to allow for evaporative cooling. Farrowing sheds were artificially lit from 0500 to 2,100. The standard farrowing space was 1.8 × 2.4 m (4.32m²), with fully slatted plastic flooring, a creep area, heated via lamp or mat for piglets, and an ad libitum feeder (Crystal Spring Hog Equipment, Ste. Agathe, Canada), two water nipples for the sow and one for the piglets. Sows were delivered 2 kg of a pre-farrow diet (12.75 MJ digestible energy (DW)/kg, 0.05 standardised ileal digestible (SID) lysine/MJ DE) at 07:00 and 15:00 prior to farrowing, with ad libitum access to a standard lactation diet (13.9 MJ digestible energy (DW)/kg, 0.6 standardised ileal digestible (SID) lysine/MJ DE) after farrowing. All sows were provided with a hessian bag (43 × 75 cm, Daish Irrigation and Fodder, Murray Bridge, SA, Australia) upon entry to the farrowing house for nesting. At approximately 24 h post-farrowing, piglets were cross-fostered where necessary within treatment. On day 2, piglets were tail docked using cauterisation, administered an iron injection (1 mL Feron 200 + B12 (200 mg/mL iron as iron dextran, 40 μg/mL cyanocobalamin), Elanco Australasia Pty Ltd., Macquarie Park, NSW, Australia), and oral coccidiostat (1 mL Baycox Piglet Coccidiocide (50 mg/mL Toltrazuril), Elanco Australasia Pty Ltd., Macquarie Park, NSW, Australia). Sow and piglets were weaned at 23.7 ± 0.23 days of lactation.

Sows were allocated to a standard farrowing crate (n = 88; FC) or a Maternity Ring (n = 83; MR) based on even parity distribution (Figure 1). The FC (Stockyard Industries, North Bendigo, VIC, Australia) was 1,783 mm in width by 2,330 mm in length, with a triangular creep area, measuring 840 mm deep and 820 mm wide. A back gate opposite the front gate provided two entry points to the pen. The crate was installed on polygrate plastic slatted flooring (Stockyard Industries, North Bendigo, VIC AU), within the pen parallel to the external dividers and measured 720 mm in width and 2,330 mm in length. The triangular piglet creep area (0.34 m²) was heated and lit via lamp (Vaucluse &APS Livestock Equipment, Inglewood, SA Australia), above a solid mat but contained no cover.

The MR was 1,800 mm in width by 2,350 mm in length, with a triangular creep area, measuring 1,070 mm deep and 920 mm wide. A back gate diagonal to the front gate provided two entry points to the pen. The ring was installed on polygrate plastic slatted flooring (Stockyard Industries, North Bendigo, VIC, Australia) and placed on a diagonal within the pen with internal dimensions being 1,160 mm in width and 2,060 mm in length and installed at a height of 250 mm from the floor. The triangular piglet creep area (0.49 m²) was heated via mat (RHS120, Stockyard Industries, North Bendigo, VIC, Australia) and fitted with a hinged lid and an LED light designed to attract piglets (29).

Production data were collected for the number of piglets born (total, born alive, and born dead) and the number of piglets weaned per sow. At 07:00 the day after farrowing, the face of each sow was inspected and allocated a facial injury score based on the spread of abrasions across three zones of the sows’ head (nose, snout and eyes/ears) as defined by Plush et al. (30). Due to the low incidence of sows presenting with a higher score (>2), this was converted to a binary trait (yes or no for facial injuries). Each sow was measured for body condition using backfat depth at the P2 site (ImaGo.S, IMV imaging, Rochester, MN, United States) on entry and exit to the farrowing house. At weaning, the presence or absence of udder damage (skin breakage) and shoulder sore was noted. Any sow or piglet displaying symptoms of illness (sow: ill thrift, vaginal discharge, mastitis, or shoulder sore; piglet: scour, meningitis, ill thrift, and physical injury) were medicated using the farms approved veterinary medication list (generally an antibiotic and non-steroidal anti-inflammatory), and these were recorded. Daily feed intake was measured for a subset of 114 sows (FC n = 58; MR n = 56). Sows were fed two times daily using a feed cart converted to scales. Each feeder was filled to a standardised volume (marked by a lip at the top of the hopper), and so the feed delivered on each day was summed to obtain the daily feed intake for each sow. Piglets were ear-tagged for individual identification, and litter size and weight were recorded after cross-fostering at 24 h and again at 18 days of age. Piglets were classified as having failed weaning if they weighed less than 3.5 kg at this time. Lactation efficiency was calculated by dividing the total feed intake by the litter weight weaned at day 18.

The following behavioural measures were analysed using footage recorded using cameras (GoPro, San Mateo, California, United States) mounted overhead farrowing accommodation. Seventeen sows (parity 2–4) were selected from each treatment; however, issues with field of view, recording duration, and image quality resulted in some footage being disregarded. The resultant sample size for each of the behavioural tests is outlined in Table 1. A single observer scored each of the behavioural tests.

Data were analysed (IBM SPSS Statistics for Windows, Version 28.0, Armonk, NY, United States) with a p-value < 0.05 considered significant, and < 0.10 a tendency. Production, performance, and behavioural data were analysed using negative binomial regression (count data), general linear mixed models (continuous data), and binary regression (yes/no data) with treatment (FC and MR) as a fixed effect and replicate (1 or 2) as a random term. Non-normally distributed data were transformed, with transformed data presented and back-transformed data in brackets. Farrowing behaviour data that were not normally distributed included birth interval mean and minimum, which were log₁₀ transformed, and the amount of time spent sitting, which was square root transformed. Data from the separation test that were not normally distributed included the duration of litter processing and the duration of the first posture change, both of which were log₁₀ transformed. Data for the anticipatory test and sow orientation on days 5 and 20 were analysed as the proportion of time a sow spent performing the behaviour, and no transformation normalised the data. Thus, these data were analysed using the non-parametric Kruskal–Wallis test. Data presented are mean ± standard error of the mean (SEM).

We had expected that sows housed in the MR treatment would show increased feed intake as has been reported previously (32), due to a greater freedom of movement leading to ease of standing and eating, and higher basal metabolic requirements. We did not find evidence of this, and in fact, crated sows were delivered significantly more feed over lactation. However, lactation efficiency tended to be higher in MR than in FC sows. Although not measured, the crated sows may have experienced increased chronic stress (33) and, as the intensity of the adrenal response to ACTH in pigs is negatively related to feed efficiency and growth rate (34, 35), acted to reduce lactation efficiency. This has sustainability outcomes other than animal welfare as if the kilogrammes of pig meat produced per sow feed consumed is improved in the Maternity Ring, the environmental footprint of the system is reduced (13).

Despite the reduced feed intake, MR sows were better able to maintain body condition than sows in crates, as measured by the change in P2 backfat thickness. FC sows lost more than 1 millimetre of backfat throughout lactation, whereas MR sows maintained condition to weaning. This maintenance of the P2 backfat may help to explain why there was a tendency for a reduced wean to the first service interval. The effect of sow body tissue mobilisation on subsequent reproduction is well documented (36–38).

Despite efforts made to reduce feed wastage, we did observe that crated sows appeared to waste more feed than those in the MR. Nine crated sows were removed from measures of feed intake due to obvious signs of excessive wastage (visible feed accruing under the slatted flooring), whilst none were removed from MR treatment (data not presented). Feeder interactions that did not involve actual consumption were measured and classified as stereotypic behaviour; however, there was no difference in stereotypies between the two treatments in the time budget analyses. Further study should be conducted into the specific feeding behaviours of crated vs. penned sows, as crated sows have no ability to turn away from the feeder, which may lead to eating for stereotypic reasons and not nutritional needs. As it is generally accepted that feed is the largest cost to pig farming enterprises, any reduction in wastage would have substantial financial benefits.

The sow’s choice of orientation is an important factor in assessing welfare, as hindered and/or enhanced expression of agency affects animals’ ability to interact with their environment (17). The MR sows were able to make a conscious choice in orientation (i.e., turn around unhindered) and so were able to actively engage with the physical and social environment beyond the degree demanded by their momentary needs. This allowed the sow to gather knowledge and enhance skills for future use in responding effectively to varied and novel challenges (39–41). The MR sows spent an average of 75% of their time facing away from the feeder, whereas in the FC treatment, this represented only 2% of observations during the time budget assessment. Therefore, when given the option a sow will make the conscious choice to spend a significant amount of time facing away from the feeder, an option not available to crated sows.

There was a reduced number of MR sows presenting with facial injuries at farrowing and so presumably less nesting behaviours were directed at hard fixtures, rather than FC sows. When confined to a barren environment, sows will perform nesting behaviours by nosing and pawing crate fixtures, becoming stereotypic in nature and resulting in injuries that can be observed on the face (42). Crated sows in this experiment were provided with hessian and so did not farrow in a barren environment; however, the number of bar biting events during farrowing and the rate of facial injury was still significantly higher. Therefore, it is the greater freedom of movement in combination with a nesting substrate that allows the sow to satisfy the need to nest build.

Almost 70% of crated sows were recorded as having udder damage at weaning, in comparison with only 12% of MR sows. Previous studies have associated increased skin lesions on the udder with lying positions (43) and flooring types (44). In the present experiment, there were no differences in lying behaviours observed during the time budget assessment, and both systems were equipped with identical plastic slatted flooring. Pedersen et al. (45) observed more teat fights between piglets and more restlessness during suckling in crated systems, and whilst teat disputes were not analysed in this experiment, this is the most likely explanation for reduced udder injuries in MR sows. Anecdotally, sows housed in the MR were able to stretch out front and back legs, allowing greater space at the udder for milk let-down events as lactation progressed and piglets grew. The position of sows during milk let-down events, and piglet disputes during feeding need to be quantified in the two systems to confirm this suggestion.

Crating sows, thereby limiting movement, has been shown to increase the incidence of ailments such as lameness and other disorders (46) (1) that require medical intervention, and so it was expected that the occurrence of such events would be lower in MR sows. We identified a trend for increased medical intervention in the MR sows in opposition to this sentiment. Six out of the 10 sows treated in the MR group came into farrowing accommodation with pre-existing ailments from gestation (one out of three in FC sows), and so the effect on medical interventions was not likely caused by the housing treatments.

Interestingly, piglet medications were fewer in the MR group. When the reasons for medical intervention were examined, the farm staff recorded more litters were treated for ‘meningitis’ in crated litters, responsible for 35% of all treatments, in comparison with 12% in the MR piglets. Meningitis is caused by Streptococcus suis, which lies dormant in the tonsils of a piglet often causing meningitis secondary to another condition, which in the pre-weaning period is most likely due to additional disease challenge or stress (47). As previously discussed, the increased udder damage observed on the FC sows may represent higher levels of teat disputes during feeding events, which could be interpreted as a stressful event for piglets and so the likelihood of meningitis symptoms. Alternatively, Nowland et al. (48) identified that piglets born to free farrowing sows ingested more colostrum which may have improved immunological status as so protection against disease. Future study should examine the welfare of piglets reared in the Maternity Ring.

There was no difference in litter size at day 1 or 18 of lactation, although numerically FC sows recorded 1.5 extra pigs at day 18. This did not translate to more pigs weaned, however, as the experimental site excluded pigs from entering the nursery facility when weighing less than 3.5 kg, with this failure rate being 1.3 pigs per litter higher in the FC sows. Piglet medications were 21% higher in crated litters which may explain this finding. The pigs that failed to wean were not removed from the system but placed on nurse sows for more lactation days to reach minimum weaning weight requirements, translating to additional labour as well as lost productive sow days and crate spaces.

The number of times a sow moved her back leg forward and performed bar biting was higher in FC sows, behaviours thought to indicate a more painful farrowing event (49). These findings are in support of those by Nowland et al. (48), who concluded that the reduced level of perceived pain may be explained by the ability of non-crated sows to move freely and become comfortable, or changes in endocrinology that results in parturition-induced hypoalgesia. Previous studies have found the level of confinement experienced by sows around farrowing to have significant physiological effects on the level of cortisol prior to farrowing (50) and the duration of farrowing (51). However, in this experiment, there were no observed differences in farrowing duration or piglet birth interval. This result is surprising as the increase in facial injury score, bar biting events, and tendency for increased stereotypical behaviours in FC sows would generally be considered indicative of increased stress around farrowing (42, 50, 52, 53). Whilst no changes in farrowing duration or intrapartum piglet deaths were observed, the behavioural observations indicate that MR sows were experiencing a less stressful farrowing process when compared to FC sows.

In early and late lactation, there was little indication of behavioural divergence between the two housing systems, which contrasts previous study that have identified shifts in both activity levels and nursing behaviour in free movement sows (54–56). This may be due to the reduced spatial footprint of the MR in comparison with other pen designs, or alternatively, the environmental conditions experienced. The experiment was conducted in naturally ventilated sheds during summer with high ambient temperatures, and observations took place during the warmest part of the day (10:00–14:00), which may reduce behavioural divergence in place of thermoregulatory comfort. In support of this, in approximately 75% of observations, the sows were scored as resting/inactive. The observation period was selected to limit the amount of disturbance during routine husbandry procedures and so stockperson interference; however, future research may benefit from investigation of sow behaviour around the busier periods in the morning and afternoon, when it is not only cooler but also when sows are more alert, active, and have more interaction with farm staff.

Sows in the MR showed a greater response to the removal of piglets during litter processing and this is not dissimilar to previous findings (57). Grimberg-Henrici et al. (58) showed that sows in confinement-free farrowing pens showed improved maternal behaviour more responsive to distress calls from their piglets. It has been suggested that the ability to perform a nose-to-nose interaction between the sow and her litter is essential in facilitating mother–offspring bonding (59, 60). Thus, the observed increase in positive interactions with piglets in the MR facilitated a stronger maternal bond and so greater reactivity when piglets were removed for processing. In support of this, the proportion of sows that performed maintenance behaviour such as eating or drinking during separation from their piglets was higher in the FC sows, and these sows had fewer contact events with piglets upon their return, indicating that crated sows were less reactive to piglet separation. This was sustained through to later lactation, where once again when observing the litters undisturbed, MR sows had more than double the contact with piglets.

The Five Domains model accepts that improved nutrition, environment, health, and behavioural conditions result in a shift in mental state of the animal towards a more positive affect (16). Mother–infant bonding is an intrinsic maternally driven process, beginning with essential nose-to-nose interactions during farrowing in the pig (Gunlach et al., 1968) (59, 60). External circumstances that enable social species to engage fully in bonding, allogrooming, play, and sexual activities have been suggested to give rise to positive affects, contributing to enhanced welfare (17). Affects that are potentially associated with aspects of bond affirmation and maternal care are of particular interest and include feeling engaged, affectionately sociable, and maternally rewarded (61–63). We suggest that MR sows experienced a more positive affect due to the increased level of interaction with piglets that was reported both during and after the sow separation test conducted during early lactation and time budget assessment occurring in late lactation.

Variation in defence cascade response is a potential indicator of affective valence and hence welfare in pigs (23). Two startle tests were conducted whilst the sows were in the farrowing accommodation. As expected, the first startle test revealed no differences in behavioural response, as the sows had not been in the accommodation long enough to alter their welfare state, therefore acting as a baseline measurement. In later lactation, sows housed in the MR displayed a lower startle score at every test, and by the final test, the score was almost 0 whilst those in an FC still recorded a score of almost two. The ability of the MR sows to turn around and investigate the environment throughout lactation possibly allowed them to gather knowledge and enhance their skills for future use in responding effectively to varied and novel stimuli (39–41). Anecdotally, it was observed that after the first aversive noise of the test, MR sows would startle and turn to investigate the source. The ability to turn allowed these sows to investigate the environment for the perceived threat, after which they would return to their normal behaviour and show attenuated responses to the remaining two aversive noises. Maternity Ring sows have a better control (and possibly agency) in their environment (as being able to assess a potential threat) and thus do not have an exacerbated response to a startling stimulus.

In the current experiment, a feed cart was presented to MR and FC sows without feed delivery with the hope of eliciting increased anticipatory behavioural patterns. Feed delivery, when restricted feeding is applied, will elicit anticipatory behaviour due to the biological needs of the sow (64). However, in the current experiment sows were fed at high levels to induce satiety prior to farrowing and ad libitum during lactation and so were not restrict-fed. If the sows were restrict-fed, this may have had adverse effects both with regard to the affect of the sows, and the impact on other recorded measures. Future study should involve the training of sows to receive a reward signalled with a cue, with anticipation measured in response to this delivery. This would have constituted a positive event, unrelated to unmet biological needs.