Judgement bias of group housed gestating sows predicted by behavioral traits, but not physical measures of welfare.

Judgement bias testing has emerged as a potential tool for assessing affective states in animals. Researchers infer an animal's affective state based on an animal's response to an ambiguous stimulus that is intermediate to both the rewarded and punished conditioned stimuli. Animals can be classified as "optimistic" or having a positive affective state if the animal displays behaviors that suggest an increased expectation of reward in the face of ambiguous stimuli. Alternatively, animals can be classified "pessimistic" or having a negative affective state if the animal displays behaviors that suggest an increased expectation of punishment in the face of ambiguous stimuli. Recent reports in multiple species question what factors influence performance in judgement bias testing, and which may allow for erroneous conclusions regarding individual affective state. In order to better understand this concern, 25 female swine were subjected to behavioral assessments at critical rearing stages to determine response variability. These same individuals were then assessed for physical measures of welfare and judgement bias using the "go/no-go" task as breeding adults. Sows which were more aggressive approached the ambiguous, but not the positive, stimulus significantly faster than others. Both optimistic and pessimistic biases were observed despite all sows living in enriched housing, and, sows with more positive physical welfare measures (fewer skin lesions and healthy body condition) did not exhibit more optimistic judgement biases. Our data demonstrate that behavior traits, such as aggressiveness, can affect a sow's performance in a judgement bias test, while measures of physical health did not. We suggest that individual differences in behavior (e.g., bold-aggressive behavioral syndrome, or, proactive coping style) generate different emotional responses and can contribute to the animal's overall affective state more so than physical ailment. Our findings highlight the complexity of how different factors impact an animal's overall affective state and support the need for complementary measures in future JBT studies, including personality assessment.

Introduction

Due to both government legislation and consumer demands, limitations have been established on the use of individual stalls to house gestating gilts and sows in many developed countries, including the EU [1], Australia, New Zealand, Canada, South Africa and ten states in the United States [2,3]. Group housing sows in pens for the majority of their gestation period allows for a greater freedom of movement, more opportunities to express species-specific behavior, and a more complex social environment. On the other hand, it can also lead to increases in injuries and lameness caused by social aggression [4,5]. Researchers often assess the welfare of individual sows in a pen by using physical metrics of poor welfare (i.e., skin lesions, lameness, or body condition), with little attention being given to the subjective psychological experience of the sows (i.e., pain/pleasure, affective state). It is particularly imperative to examine this latter issue in swine breeding animals (e.g., parent stock) as they live much longer lives than market hogs and could potentially suffer long-term negative emotional states.

The evaluation of the emotional states in non-human animals rely on physiological and behavioral indicators of arousal, as well as measures of conditioned responses to infer pain or pleasure experience (e.g., motivation and aversion testing) [6]. This evaluation can be expanded to include the cognitive component of an emotional state, such as, the impact of emotions on information processing (e.g., attention, memory, and judgement/appraisal) [7,8]. Research on human and non-human animals indicates that negative mental states, such as anxiety or depression, can induce pessimistic appraisals of ambiguous stimuli (i.e., expectation of punishment, or, lack of reward) [9,10]. For example, rats, piglets, and starlings in barren, crowded, or unpredictable housing conditions will display pessimistic cognitive bias in their appraisal of ambiguous stimuli, and thus, are inferred to experience more negative affective states [11-13]. While this assessment of ambiguous cue interpretation allows researchers to infer how an individual’s emotional state may influence information processing, it typically does not consider the influence of individual differences in behavioral response (i.e., personality) or individual sensitivity to reinforcer [14].

Personality can be described as characteristics of behavioral reactions such as intensity, speed, response threshold, latency and recovery time, which are most clearly expressed in novel and unpredictable situations, and are relatively stable throughout an individual’s development [15]. Given that personality is most clearly expressed in novel and unpredictable situations, behavioral tests are constructed to elicit high-arousal responses. Individual differences in personality may influence immunity strength [16], the performance of abnormal or stereotypic behaviors [17,18], and the expression of pain [19]. New research suggests that an animal’s personality can also influence the way in which stimuli are processed and evaluated; in other words, personality can shape judgement biases [20-27]. For example, proactive (exploratory and bold) grower pigs are more likely to display optimistic judgement biases than reactive (non-exploratory and cautious) grower pigs [20], dogs which receive higher traits scores for sociability, excitability and non-social-fear display more optimistic judgement biases than more fearful, less social dogs [21], and, neurotic orange-winged amazon parrots show a greater attention bias for environmental stimuli [22].

Given that breeding sows are the longest living population of swine in a production setting, it is important to understand the causes, mechanisms, and indicators of long-term negative emotional states to prevent possible suffering. As reviewed in a comprehensive book on the welfare of production swine [28], there is considerable attention in the literature given to the health and welfare of gestating and lactating sows as inferred from biological measurements and behavioral indicators. In contrast, to our knowledge, there has been only one study which utilizes cognitive methods (i.e., judgement bias paradigm) to infer affective state in gestating sows [29]. There are far more studies evaluating mood among younger animals, such as the investigation of stocking density impacting performance in judgment bias test among 14-week-old pigs [30], or the use of associative learning among 12-week-old pigs to evaluate emotional contagion [31]. In addition, while there have been approximately 48 studies on coping styles or personality traits among swine [32], less than a dozen specifically address consistent individual differences among breeding sows [33-38]. Given that there is limited information on the use and validity of judgement bias testing to evaluate affective states in sexually-mature sows, there is a need to investigate possible influences on performance in this cognitive task.

The aim of the present study is to investigate whether consistent individual differences in behavior of breeding sows predicts performance in a spatial discrimination judgement bias task. Based on recent reports linking animal personality, or coping style, to performance in a cognitive bias test, we predict that one or more behavioral traits evaluated will predict performance (e.g., response latency) in a spatial discrimination task. In addition, if judgement bias reflects the affective state of an animal (i.e., biological health and mental state), then it is important to test how physical measurements of welfare (i.e., skin lesions, body condition) also could impact performance in judgement bias task and contribute to mental well-being.

Methods

Animals and housing

Animals observed in this study were born and reared at the Swine Teaching and Research Center located at the School of Veterinary Medicine, University of Pennsylvania. At this research facility, approximately 120 to 130 gestating sows are housed in a dynamic, multiparous group pen and fed via electronic sow feeding stations (ESF, Schauer Agrotronic Compident 7). The sow pen allows for 2.4 m2 per sow, and includes nine 3.0 x 2.1 m concrete lying areas, a 4.0 x 17.0 m straw bedded area and a total of 183.9 m2 of slatted flooring. These sows farrow (give birth) in farrowing pens with dimensions of 2.1 x 2.0 m. Each farrowing pen was equipped with a hinge-sided farrowing crate that initially confined the sow immediately post-partum, but could be opened to provide the sow additional space and freedom of movement as the pigs matured. Piglets born to these sows are weaned to ~0.4 m2 /animal in groups of 18 piglets. At ~3.5 months of age, the replacement gilts are housed in pens of four gilts at ~0.7 m2 /animal until the age of ~6.5 months when they were selected to enter the breeding herd. For the present study, 25 replacement gilts (PIC 1050, Landrace-Yorkshire crossbreed) were placed in behavioral experiments at ages 5 weeks, after first farrowing (10 to 17 months) and during second gestation cycle (11 to 18 months). Each animal was trained and tested for judgement bias approximately two weeks after mixing into large gestational sow pen as primiparous sows. All procedures were performed in accordance to the approved guidelines by the University of Pennsylvania’s Institutional Animal Care and Use Committee, protocol #804656. Data collection methods for the restraint test, response to litter restraint test, and, behaviors recorded at pen mixing have previously been described in our study of the ontogeny of personality trait assessment in production sows [38]. These particular tests were intentionally designed to incorporate routine animal husbandry procedures in order to adapt for on-farm application and to elicit high-arousal and unpredictable situations. A timeline of data collection schedule is summarized in Fig 1.

Fig 1

Timeline of data collection.

Data collected at 5 weeks old (Restraint Test), 24 hours post-farrow (Litter Restraint Test), and, 1 hour (Social Interaction Test) and 2 weeks (Judgement Bias Test) post-mixing into group pen, as well how data collection corresponds to the life cycle of the sow.

Data collection

Restraint test

At 5 weeks of age, each piglet (n = 25) was evaluated for behavioral response to being held in the arms of an experimenter for 30 s. This experiment was intended to mimic the routine handling procedures on the farm, such as during piglet processing. Researchers use similar restraint tests (i.e. back-test) as a non-social measure of stress response in order to classify piglets as ‘high-resistant/proactive’ or ‘low resistant/reactive’ based on pre-determined cut points in the behaviors that are measured. The same experimenter performed the restraint test for all piglets. This experimenter stood in one corner of the pen for 60 s to allow for the piglets to acclimate to the presence of a human. The piglets were tested in a semi-random order, with the experimenter bending down to pick up one piglet at a time, often choosing the piglet that is in closest proximity in order to reduce the movement of the piglets in the pen. The experimenter lifted each piglet by a back leg, while supporting the piglet’s chest with their other hand. The test began once the experimenter was standing still while supporting the piglet’s body with both hands, with one hand cradling the torso and the other hand supporting the hips, and resting a side of the piglet body on the human’s chest. The experimenter held the piglet in a prone position firmly and did not move their hands or change the piglet’s position during the 30 s experiment. Each piglet was given a score based on the following scale: 1: No vocal response or physical movement; 2: Less than 5 seconds of soft vocalizations (grunts, squeaks) and/or physical thrashing; 3: 5 to 10 seconds of squealing, grunting and/or physical thrashing; 4: 10 to 20 seconds of high-arousal vocal response (squealing) and/or physical thrashing; 5: More than 20 seconds of high-arousal vocal response (squealing) and/or physical thrashing. The primary experimenter holding the piglet counted in their head continuously (e.g., “one-one-thousand, two-one-thousand, etc.) only when the piglet vocalized or struggled in their arms. A secondary experimenter observing the test recorded the duration of physical struggling and vocalizations continuously using a stopwatch. The scores of each piglet (1–5) were compared between the two experimenters in order to assess the inter-observer reliability (Cohen’s kappa > 0.85). Each piglet was returned to the pen after receiving a single mark on their back using a non-toxic livestock marker to indicate that the animal had been tested.

Response during litter restraint test

Each sow’s reaction toward a human handling her first litter during piglet processing was recorded 24 h post-farrowing (sow age range = 10 to 17 months) The piglet processing procedure involved each piglet being removed from the farrowing pen and being placed in a 1.4 x 0.5 x 0.9 m transport cart with straw bedding (QC Supply; Schuyler, NE, USA) to receive an intramuscular injection of iron dextran. The sow was kept in a closed farrowing crate (0.6 x 1.7 m) while the piglets were removed. The test began when the first piglet was removed from the pen and placed in the transport cart, and the test was complete after the last piglet was returned to the pen and the human exited the farrowing pen, with a test duration range of 5.5 to 8.7 min (median = 6 min). Each sow was given a score based on the behaviors and vocalizations she exhibited throughout the piglet processing (Table 1).

Table 1

Operational definition of each score given to sows during the litter handle test when their piglets were removed from pen during processing.

Score	Operational Definition
1	Sow does not orient to human, does not vocalize, and remains recumbent during test.
2	Sow takes more than 5 s to change posture while remaining recumbent to orient toward human, and produces less than 5 soft vocalizations (grunts).
3	Sow takes more than 5 s to change posture from recumbent to sitting posture, orients toward human, and produces more than 5 soft vocalizations (grunts).
4	Sow takes more than 5 s to change posture from recumbent to standing and produces mixture of soft and loud vocalizations (grunts and barks).
5	Sow takes less than 5 s to change posture from recumbent to standing, produces loud vocalizations (barks), and attempts to bite or lunge at human in the farrowing pen.

Social interaction in gestation pen

The primiparous sows were housed in standard breeding stalls (2.0 x 0.6 m) for seven days after weaning (sow age = 11 to 18 months). Each primiparous sow was then moved from the breeding stalls to the dynamic sow pen, containing 120 to 130 multiparous sows, in batches of 8 to 10 sows approximately 4 to 5 days post-breeding. Each sow was filmed for 1 hour following introduction to gestation pen given that previous studies have reported most aggressive encounters occur at the beginning of group formation [39,40].

Video data was coded for the frequency, duration, and outcomes (displace other or retreat) of fights at mixing of focal sows into the group pen. The duration of fighting was recorded as the start and stop of parallel-pressing; defined as the sows standing side-by-side and pushing the shoulders against each other, with more than three bites initiated or received by the focal sow. Additional behaviors coded included the frequency of aggressive behaviors (bites, chasing) initiated and received and the number of nose-other investigations initiated (Table 2). Video data were analyzed using a continuous focal animal behavior sampling method (The Observer XT 11; Noldus Information Technology, Wageningen, The Netherlands). Video data were coded by three observers, with high inter-observer reliability across all behaviors (Intraclass correlation coefficient > 0.8).

Table 2

Ethogram of behaviors coded during the 1-hour post-introduction to an unfamiliar, dynamic gestation pen 5 days following weaning and breeding.

Behavior Event	Operational Definition
Bite	Sows uses teeth to clamp down or scrape another sow’s hear or body.
Chase	Sow purses another sow at a fast pace, trying to reduce the distance between the dyad.
Displace	Sow forces another sow to move away from current location. May or may not be at the end of a fight.
Fight	A dyad of sows engages in reciprocal hits and bites while pressing shoulders against each other (face-to-face).
Hit	Sow uses snout or side of head to hit another sow’s face or body with force.
Lie down	Sow lies lateral or sternal recumbent position with eyes open or closed and not nosing the environment or another sow.
Nose other	Sow places snout near (<4 cm) the snout, head, body or limbs or another sow.
Retreat	Sow moves away from another sow, at a walking or running pace. May or may not be at the end of a fight.

Physical measures of welfare

Lesion, lameness, and body condition scores were assessed 3 days post-mixing. Lesion scores were evaluated using a previously designed scale for sows [37]. This scale identified lesion quality (A: Thin (<2mm wide) and shallow; epidermis may be broken with mild surface inflammation; B: Moderately deep cuts (2 to 4mm wide); epidermal and dermal layers may be broken with moderate inflammation; C: Occurrence of wide (>4mm) or deep cuts; dermal layers broken with severe inflammation), and lesion quantity (1:≤5 cuts; 2: 6 to 10 cuts; 3: 11 to 20 cuts; 4: 21 to 30 cuts; 5: ≥31 cuts). Each sow received a quality and a quantity score for the following body regions: anterior (cranial to the caudal aspect of the shoulder), side (between the caudal shoulder and cranial hip), and, posterior (from the cranial hip caudally). For analysis, each quality score was transformed in to a number (A = 1, B = 2, C = 3) and was multiplied by the quantity score (1 to 5). Therefore, the minimum score for each region was 1, while the maximum score for each region was 15.

Lameness scoring was based on the Zinpro Feet First ® (Eden Prairie, MN, USA) scale [41] of 0: sow moves easily with little inducement with no signs of discomfort; 1: sow moves relatively easy, but visible signs of lameness are apparent in at least one leg; 2: lameness is involved in one or more limbs, and the sow exhibits compensatory behaviors such as dipping her head or arching her back; and 3: there is a real reluctance to walk and bear weight on one or more legs, and, it is difficult to move her from place to place on the farm. Body condition scoring was based on the industry standard [42] evaluation of back-fat thickness and prominence of hipbones and spine using the following scale: 1: emaciated; 2: thin; 3: ideal; 4: fat; and 5: overly fat.

Judgement bias testing

Each sow was trained and tested for a judgement bias 2 weeks post-introduction to the large gestation pen. A simplified “go/no-go” task with a single ambiguous cue was designed to facilitate training and throughput to ensure that we were able to test this cohort of animals at a similar age and stage of gestation. Experimental set-up, training and testing of the behavioral task was previously described in our study of judgement bias assessment in a separate cohort of multiparous sows [29]. For both the positive and negative stimuli, a single 25.4 cm in diameter metal pet food bowl was filled with 200 g of corn-soy meal and placed in the far left or right corner of a 2.4 x 8.3 m experimental field (Fig 2). The first 3.7 m of this field was walled with steel gating, while the remainder of the field was walled with cement. The designation of positive versus negative stimulus location (left or right) was randomized for each sow. Using operant conditioning of paired associations, sows were trained to distinguish between the two visual stimuli based on the location of the food bowl. Only one bowl was present in the field during each trial, with a primary researcher standing behind the bowl, holding the red flag behind their back, in order to keep a consistent protocol. This primary researcher kept their gaze forward and did not interact with the sow. A second researcher acted as the animal wrangler; releasing the animal from the starting gate and retrieving the animal from the experimental area after each trial. Animals were trained at approximately the same time each day (1000–1400) for a maximum 12 trials or a maximum time of 30 minutes. The order in which the animals were trained and tested was dependent upon the time of their last meal, in order to ensure that each animal was trained at least 4 hours after eating their daily feed at the ESF stations.

Fig 2

Schematic representation of judgement bias experiment.

Sow was held behind gate at starting line (long dashed line). During training, the conditioned feed bowls were place in either far left or right location of the arena. Only one feed bowl was present at a time. Testing trials placed feed bowl in the ambiguous stimulus position (?), equidistant from rewarded (+) and punished (-) locations. Approach was defined as the latency for the sow to leave the starting gate and cross the goal line (short dashed line) with at least one front limb.

On the first day of training, the sows were conditioned to approach the bowl in the positive location over a series of ten trials. Approach was defined as the latency for the sow to leave the starting gate and cross the goal line of a 1 x 1 m area containing the feed bowl with at least one front limb. The distance from the starting gate to the goal line was 7.3 m. If the sow approached the positive stimulus, she was allowed to consume the corn-soy feed for 10 s. The second day of training began with a positive trial for reference, followed by a minimum of 4 negative trials, and ending with one positive trial. If the sow approached the feed bowl in the negative stimulus location, the standing researcher would wave a red flag over the bowl for 20 s, preventing the sow from consuming the feed. The third and fourth training day began and ended with a positive trial, with eight randomized positive or negative trials in between. Sows were considered trained if they approached the positive stimulus within 20 s for 100% of the positive trials and if they did not approach the negative stimulus at least 50 s for at least 80% of the negative trials. The speed criteria were chosen based on average walking speed to approach bowl from starting gate for a sow. The minimal successful no-go response was determined empirically based on observation of the sows and distribution of the data [35]. The experiment day included three trials: a positive trial, a negative trial, and an ambiguous trial. In this ambiguous trial, the feed bowl was placed in the middle of the experimental field, equidistant to the positive and negative stimuli (Fig 2). The latency of approach to this ambiguous stimulus was recorded, with a maximum time of 60 s (i.e., no-go). Additional probe cues, placed between the conditioned stimuli and the ambiguous bowl position, were not used in this experiment as the pregnant sows were unable to perform multiple trials within a 30-minute time frame. Given the physical limitations of the animals, and that an elongated testing period to include probe cues would find the sows in heavier body conditions due to growing gestation, we relied on behavioral response to a single ambiguous cue.

Data analysis

Principal component analysis

The eleven behaviors recorded during pen introduction were transformed into composite traits using principal component analysis (PCA) with a Varimax rotation (IBM SPSS Statistics 24, Armonk, NY, USA). Prior to PCA, the adequacy of the data was examined, ensuring that Bartlett’s test of sphericity was significant (P<0.001) and the Kaiser-Meyer-Olkin measure was high (KMO = 0.64). Components which had eigenvalues greater than 1 were retained for interpretation. Behaviors were grouped based on a loading cutoff of 0.6. Sows received a single component score for each trait using the least squares regression approach. Significant relationships among the variables were assessed using a non-parametric Spearman’s rank-order correlation (P<0.05).

Modeling of study outcomes

A linear mixed-effects model with restricted maximum likelihood was implemented in R using the lme4 package [43]. We analyzed individual animal single trial latencies as a continuous outcome variable, with latency being defined as the time to approach the specific stimulus cue. The initial model included as fixed effects: probe location, behavioral trait scores from the principal component analysis of social interaction, skin lesions scores, restraint test scores, and response to litter restraint test, as well as their interactions with probe location. Fixed effects were dropped if they did not influence model fit as assessed using ANOVA. Lesions scores and the submissive and active trait scores were dropped from the final models for this reason. Sow identity was included as a random effect. To better understand interactions in the models, post hoc linear models were fitted to examine how the different variables above influence the time to approach specific probe locations (positive, negative and ambiguous). Visual inspection of residual plots did not reveal any obvious deviations from homoscedasticity or normality. P-values were obtained by likelihood ratio tests of the full model with the effect in question against the model without the effect in question. Ordinal variables such as lesions scores and human interactions scores were only included in models as independent variables [44]. In these cases, model coefficients can be more difficult to interpret and as such are not presented. Median and interquartile range (IQR) were used to describe these count data whereas mean and standard error of the mean were employed for latency measures. Power analysis using power.f2.test in R revealed that with the study sample size, a significance of 0.05, and a power of 0.8 that we could detect an effect size of 0.35 or larger.

Results

PCA analysis

Three principal components which explained 79.2% of the variance were extracted from the dataset (Table 3). The first principal component had strong (>0.6) positive loadings for the following behaviors: number of bites and chases initiated, the duration of chasing another sow, the number of displacements of other sows, and the total number of fights engaged. The second principal component had strong positive loadings for the following behaviors: number of retreats from another sow, the number of bites received, and the duration of being chased by another sow. The final principal component had a strong positive loading for the number of nose-investigations of other sows, while it had a strong negative loading for the duration of lying down during pen mixing. Based on the behaviors which loaded onto each component, they were labeled: aggressive, submissive, and active/exploratory.

Table 3

Obliquely rotated principal component loadings for behaviors recorded during the 1-hour post-introduction to an unfamiliar, dynamic pen as primiparous sows.

	PC1	PC2	PC3
	Aggressive	Submissive	Active/Exploratory
Bite–initiate (#)	0.92	0.07	0.14
Chase—initiate (#)	0.87	-0.02	0.12
Chase–initiate (s)	0.80	-0.09	0.001
Displace (#)	0.72	-0.20	0.54
Fight (#)	0.86	0.15	0.14
Bite–receive (#)	0.58	0.65	-0.06
Chase—receive (#)	0.20	0.92	0.07
Retreat (#)	0.41	0.78	0.04
Lie down (s)	0.12	-0.39	-0.78
Nose other (#)	0.42	-0.19	0.73
Eigenvalue	4.3	2.1	1.5
Variance explained (%)	42.9	21.4	14.9

Rotation Method: Varimax with Kaiser Normalization. Rotation converged in 5 iterations.

Component loadings >0.6 (bold face) were clustered to create composite traits.

Judgement bias assessment

Of the 25 sows trained, only one sow did not meet the task requirements on test day and was removed from the analysis. The remaining 24 sows approached the positive cue in less than 20 sec and failed to approach the negative cue for at least 50 sec. When approaching the ambiguous cue, eleven animals exhibited a latency of less than 20 s with an average time of 9.0 +/- 1.2 sec; nine animals took longer than 50 s with an average time of 59.6 +/- 0.4 s; and 4 animals exhibited a latency of between 20 and 50 s with an average of 30.8 +/- 3.8 s (Fig 3).

Fig 3

Judgement bias outcomes vary across group housed sows.

Individual latencies for sows to approach the positive, ambiguous, and negative stimuli area displayed as well as means for animal exhibiting either an optimistic bias (ambiguous approach < 20 s), an undetermined bias (ambiguous approach, between 20-50s) or a pessimistic bias (ambiguous approach > 50 s). Both optimistic and pessimistic bias animals approach the positive and negative conditioned stimulus with similar latencies independent of their judgement bias.

Relationship between traits and judgement bias

A linear mixed effects model revealed that both the location of the stimulus (χ2 (2) = 102.83, p<0.001) and trait scores for aggression (χ2 (1) = 8.81, p = 0.003), restraint test score (χ2 (1) = 5.44, p = 0.020), and the response to litter restraint (χ2 (1) = 4.93 p = 0.027) significantly affected the latency to approach ambiguous stimulus. There were no significant effects of the submissive or active/exploratory trait scores on the latency to approach positive, ambiguous or negative stimulus. Interactions between probe location and each of these three variables were also significant (aggressive trait score: χ2 (2) = 21.40, p < 0.001; restraint score: χ2 (2) = 11.24, p = 0.004; and litter restraint score: χ2 (2) = 11.27, p = 0.004). Specifically, the approach to the ambiguous stimulus was affected by the trait scores for aggressive (F(1,22) = 11.49, p = 0.003), with more aggressive sows displaying shortened approach times to the ambiguous cues or a more optimistic judgement bias (Fig 4A). However, aggressive sows were not quicker in general as there was no significant relationship between the approach to the positive cue and the aggressive trait score (F(1,22) = 0.21, p = 0.648; Fig 4B). Similarly, the approach to the ambiguous stimulus also was affected by the restraint test score (F(1,22) = 5.74, p = 0.026), with sows that exhibited more resistance to restraint as a piglet displaying more optimistic judgement bias. These sows also were not quicker in general as there was no significant relationship between the approach to the positive cue and the restraint test score (F(1,22) = 1.57, p = 0.223). The approach to the ambiguous stimulus also was affected by the response to litter restraint scores (F(1,22) = 5.55, p = 0.028), with sows that exhibited a stronger response to the restraint of their litter displayed a more optimistic judgement bias. These sows also were not quicker in general as there was no significant relationship between the approach to the positive cue and the response to restraint litter (F(1,22) = 0.00, p = 0.951).

Fig 4

Personality influences judgement bias of gestating sows.

(A) Significant relationship between the trait scores for aggressive/dominant and the latency to approach the ambiguous cue (p = 0.003), with more aggressive/dominant sows approaching the ambiguous cue faster; indicative of a positive judgement bias. (B) Aggressive/dominant animals, however, do not approach the positive cue faster than less aggressive animals (p = 0.65).

Physical measures of welfare

Skin lesions acquired following the introduction of study animals into an unfamiliar group of sows ranged from a score of 1 to 15 and were most common in the anterior region (median = 12; IQR = 9), intermediate for side lesions (median = 6; IQR = 5), and least common in the posterior region (median = 4; IQR = 2). A significantly positive relationship was observed between the submissive trait scores and lesion scores of anterior region 3 days post-mixing (F(1,22) = 7.24, p = 0.013) (Fig 5A). However, no significant relationship was observed between any of the other regional lesion scores and the submissive trait or any other lesion scores and the other two personality traits (aggressive/dominant or active/exploratory). There also was no significant relationship between the lesions scores of each body region and judgement bias (Fig 5B). The latency to the ambiguous cue was not significantly related to an animal’s anterior lesion score (F(1,22) = 1.54, p = 0.699), side lesion score (F(1,22) = 1.05, p = 0.316), nor posterior lesions score (F(1,22) = 0.03, p = 0.861). Furthermore, neither lameness nor body condition score impacted judgement bias. No animals were evaluated to be lame during the study (i.e., a lameness score ≥ 2), and thus, lameness could not account for pessimistic biases observed in nine sows. Similarly, only 3 of the 24 of the animals at the time of judgement bias testing had a sub-normal body condition (BCS = 2) and the latency to the ambiguous cue was not significantly related to an animal’s body condition score (F(1,22) = 0.67, p = 0.423).

Fig 5

Physical measures of animal welfare do not predict with judgement bias of gestating sows.

(A) Significant relationship between anterior skin lesion scores and submissive trait scores (p = 0.013) as animals with more severe skin lesions had higher submissive trait scores. (B) Animals with less severe anterior lesions scores, and thus better physical welfare, did not exhibit faster approaches to the ambiguous cue as no significant relationship was observed between lesion scores and latency to approach the ambiguous cue (p = 0.699).

Discussion

Optimistic, pessimistic, and undetermined judgement bias were observed among a cohort of 25 out of 130 gestating sows that shared similar genetics and rearing conditions, and lived in the same pen. We did not find that animals with increased lesion severity displayed more pessimistic judgement biases. For this study cohort, variation in body condition and lameness was too small for meaningful analyses in regards to performance in judgment bias test. Instead, head and shoulder lesion severity following introduction to the gestation pen was significantly correlated to the personality trait of submissive. In our previous work, sows which had higher aggressive trait scores were evaluated to have more severe lesions in the anterior region [37]. This was not found in the present study and may be the result of a smaller sample size, but it does indicate that the variability of lesions acquired for group housed sows is linked to the trait of submissive/non-aggressive and dominant/aggressive. Previous reports have also found that higher lesions scores are not necessarily indicative of a more aggressive sow, but instead, may signify something more about the aggression level of the pen [45]. Furthermore, we found that sows which were determined to be more resistant during restraint as a piglet, more aggressive toward humans as an adult when their piglets were handled, and more aggressive to conspecifics during social mixing, were more likely to display optimistic biases as compared to non-responsive gilts and non-aggressive sows. Taken together, these results suggest that the judgement bias paradigm used in the present study more likely was evaluating a sow’s response to risk as it relates to consistent individual differences in behavior (i.e., personality) and raises the question of how personality attributes contribute to the subjective mood state (i.e., mental welfare) of the sow. Further research is needed to evaluate if this spatial judgment bias task is capturing emotional differences in temporary appraisal bias (i.e. mood state) or a more permanent factor of the animal’s character (i.e., affective or coping style) [46].

The spatial learning task has previously been used to assess judgement bias in younger pigs based on variation in housing enrichment [13], following social isolation [47], and following a change to the stocking density [30]. Unfortunately, the use of such cognitive methods to evaluate affective states of mature swine, like breeding sows and boars, is not as well-established in the literature. These breeder animals live the longest lives of any animals on a swine farm and have the potential to endure the most suffering from negative affective states. Current welfare assessments for breeding sows focus on physical ailments (e.g., lameness, body condition and skin lesions), as well as being the target of aggressive behaviors (e.g., bites, scratches, displacement). However, given that recent work in dairy heifer welfare has reported a lack of correlation between behavioral indicators of anhedonia and the number of agonistic interactions received during regrouping [48], there is a need to carefully critique our assumptions of how welfare is experienced by the animal. If we are using judgement bias testing as an indicator of subjective mental welfare of animals, we also need to better understand how, and if, the degree and duration of physical comprise contribute to an animal’s psychological welfare state.

Previous research evaluating comprehensive welfare assessments of stalled versus free-ranging horses found that individuals which live in more restrictive, non-social environments not only displayed the highest levels of behavioral and chronic health-related problems, but they also showed more hesitation to ambiguous probes (i.e., negative judgment biases) [49]. In our study, however, there was not a significant difference in the location, quantity, or quality of lesions between sows with optimistic biases and those with pessimistic biases. An a prioiri assumption would be that sows with more severe lesions would display more pessimistic biases; given that the lesions could result in a negative state like pain (or be caused by engaging in negative social behaviors and result in a fear or anxiety state). This relationship, however, was not found, and thus our data do not support an impact of physical measurements of welfare (i.e., skin lesions, body condition) on judgement bias in sows. Given the two-week delay between the acquisition of lesions at mixing and the judgement bias testing, the time horizon over which these lesions could negatively impact the sow’s mood from pain, fear or anxiety maybe relatively short-lived compared to the several weeks that it takes for the skin lesions to heal [50].

In behavioral ecology, the relationship between speed-accuracy tradeoffs and personality differences in risk-reward tradeoffs are examined as impacting individual fitness through fast or slow response to environmental change [51,52]. Similar to the bold-aggressive behavioral syndrome, which is represented by fast-responding animals that take large risks for high reward (with poor accuracy in gaining reward), animals with proactive coping styles are also bold (e.g., shorter approach/attack latencies), aggressive, and rely on routine behavior (i.e., low flexibility in response to change) to cope with challenges in their environment [53]. Reactive individuals are more cautious (e.g., longer approach/attack latencies), less aggressive and are more adaptive to change by developing different behavioral patterns. In keeping with this theme, we have previously reported that dominant sows exhibit shorter latencies to approach an ambiguous stimulus [29]. In order to improve upon the methods conducted in this study, it is recommended that sows are placed in multiple contexts (e.g., appetitive-consummatory behavior, social behavior, human interaction) and each test is repeated to establish trait consistency.

Several possible limitations to our study warrant discussion. One potential limitation of this study is that we used a locomotion-based go/no-go task, which prevented severely lame sows from participating in the trials. It is thus possible that more negative states of physical welfare, such as resulting from more severe lameness, might have a greater impact on the sow’s subjective mental experience, and thus, influence judgement bias. Another possible limitation of this study was that it was only sufficiently powered to detect large effect sizes. Thus, the failure of this study to find a relationship between physical welfare and judgement bias could also be explained by there being only a small contribution of physical measures of welfare to affective state as such we would not have observed it. Another possible limitation of this study is that only one ambiguous probe cue was used in the go/no-go paradigm, instead of a more common three intermediate probe cues [8,9]. The operant conditioning schedule for this study was intentionally limited to one week in order to reduce the chance for fatigue or physiological factors related to pregnancy to influence behavior in the locomotion task. Each sow was trained for a maximum of 30 minutes per day, as previous research reports that gestating sow spend over 70% of their day lying down in rest or sleep state. In addition, commercial breeding sows are fed approximately 30% of ad-libitum consumption per day, which meets nutritional needs, but does not result in satiety [54]. Therefore, our sows were chronically hungry. Given that the reward for our study was access to their soy-corn diet, presenting multiple ambiguous trials would be difficult to interpret, as all sows were approaching a bowl if it was not punished (i.e., an unrewarded bowl was still reinforcing for these hungry sows). Future work could attempt to either pre-feed sows prior to testing and training, but research has shown that approximately 6.8 kg is need for a sow to be sated, or, the reward could be non-food variety, such as access to straw to enrichment. A final possible limitation is in how the data was collected. In the latency to approach the ambiguous cue, the trial was ended if after 60 seconds the sow failed to approach. This data has been described ‘continuous but censored’ and the concern is that these censored latencies might underestimate pessimistic bias in quantitative model [55]. However, the results suggest that the censored data does not influence our interpretation of the findings. In Fig 4A, longer latencies for the animals with pessimistic biases only would have accentuated the inverse relationship between latency and trait score. In Fig 5B, the animals with censored latencies were observed across the full range of lesion scores consistent with no relationship between latency and skin lesion scores.

An additional caveat to interpreting animal behavior in judgement bias tests is that bold, exploratory animals may appear to perform better at the go/no-go task simply because they are quicker to respond. For example, mice which display fast exploration of the open areas in an open field test (i.e., bold/exploratory trait) show superior performance in a variety of spatial and discrimination tasks than non-exploratory or cautious mice [56]. It can be argued that animals which are less prone to novelty seeking, and are slow to explore their environments, may appear to be more hesitant in the go/no-go task; and thus, are assumed to experience more negative affective states than animals which are quick to process and respond to environmental stimuli. This very notion has been reported in a study with laboratory mice, with more exploratory, less neophobic mice taking time to investigate novel, ambiguous arms than reach the reward end [57]. An inherent reason for using judgement bias testing to assess affective states in animals is to investigate how emotions impact an organism’s judgement of risk in their environment (through approach-avoid behavior). However, without overt manipulation of an animal’s subjective mental state (e.g., painful husbandry practices, as seen in negative judgement bias post-horn disbudding in dairy calves [58]), or control for the pleasure versus aversive experience of the reward and punishment (e.g., 6 versus 1 apple slice for grizzly bear judgement bias test [59]), it is difficult to ascertain the reason for an animal to hesitate approaching or responding to an unknown stimulus; it could be caused by a short-term affective state, or a long-term affective style (i.e., personality trait [60]).

Conclusions

In our longitudinal study of personality development among a cohort of domestic sows, individuals which were determined to be more resistant during restraint as a piglet were more aggressive as an adult during social mixing and when their own piglets were handled during processing [1]. The results from the current study indicate that these aggressive individuals were also more likely to display optimistic biases as compared to non-aggressive, non-responsive sows. This suggests that animal personality may impact cognitive processes, perhaps through different emotional states, and influence a sow’s appraisal of unknown stimuli in the environment. Although the individuals of this study were housed in same enriched environment, both optimistic and pessimistic biases were observed and these judgement bias outcomes did not correlate with contemporaneous physical measures of animal welfare. Based on the results of this study and reconciling a growing body of evidence in the literature, we suggest that individual differences in behavioral traits generate different emotional responses that can contribute to the animal’s overall affective state. This work also supports variable contributions to affective state given the poor correlation of judgement bias with transient physical compromise. Our findings highlight the complexity of how different factors impact an animal’s overall affective state and argues the need for additional, complementary measures in future JBT studies on group-housed gestating sows, including personality assessment.

(XLSX)

Click here for additional data file.

22 Sep 2021

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Miquel Llorente, PhD

Academic Editor

PLOS ONE

Journal Requirements

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2, We note that the grant information you provided in the ‘Funding Information’ and ‘Financial Disclosure’ sections do not match.

When you resubmit, please ensure that you provide the correct grant numbers for the awards you received for your study in the ‘Funding Information’ section.

3. Thank you for stating the following in the Acknowledgments Section of your manuscript:

“The authors would like to thank the summer research interns of the Swine Research and Teaching Center for their assistance in data collection. This work was supported in part by Kraft Foods Oscar Mayer, Pig Improvement Company, Pennsylvania Pork Producers Council, Pennsylvania Soybean Board, Pennsylvania Department of Agriculture, and the American Society for the Prevention of Cruelty to Animals”

We note that you have provided additional information within the Acknowledgements Section that is not currently declared in your Funding Statement. Please note that funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

“This work was supported in part by Kraft Foods Oscar Mayer, Pig Improvement Company, Pennsylvania Pork Producers Council, Pennsylvania Soybean Board, Pennsylvania Department of Agriculture, and the American Society for the Prevention of Cruelty to Animals. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript”

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This is an important subject, to challenge the recent tendency to over-interpret results from cognitive bias paradigms as de facto evidence of good or bad welfare. It would be nice to see consideration of alternative explanations for cognitive bias results become the norm.

Reviewer #2: This paper aims at exploring what factors influence performance in judgement bias testing in group-housed sows with an emphasis on personality traits and measures of physical aggressions.

This is an important area of research given the need for a better understanding of affective states of farmed animals and how it can successfully relate to other welfare measures.

I think the study has merits, but I feel it could really be improved. I spent quite some time making both general and detailed comments that I hope will help the authors.

General comments

• The literature considered in this manuscript is rather incomplete as work on other species (dairy calves and rats) has not been integrated. I would strongly recommend including what has been done in other species given that only a few studies have explored individual variation in response to judgment bias. For now, the authors do not really refer to studies exploring this link but rather mention their existence. I think the integration of these studies will provide more context and should help the authors better communicate their results, especially in terms of how moods can be considered both as reflecting emotional states and/or personality differences.

• The rationale behind the research question could be improved. For now, the reader is left with limited information on why the authors did this study and why it is important.

• I’d like the authors to discuss more some limitations of their study. For instance, the authors used only one ambiguous cue presented only once, which is at odds with the literature. The authors need to explain why they made that choice and more effectively explain how this could have affected their results (i.e., animals may be more incline to react to ambiguous cues as novel element in their environment on the first exposure vs. on repeated exposures).

• Many scientists would argue that personality traits need to be assessed over time and contexts. In addition, tests were not all standardized (e.g., the social interaction test was more a regrouping experiment), which is a limitation of the study. I would recommend to keep the concept of personality for the discussion but I may be convinced otherwise if the authors clearly stress the consistency of their outcome variables.

• What is the point of the JBI? I would remove this part as the authors do not seem to use it for further analyses.

Abstract

L21 states?

L22 I feel the authors should give more information to the reader so that it is easier to grasp how JBTs work. I would add elements explaining pessimistic and optimistic responses.

L26 Maybe better to split this sentence?

L36 I would tone down the conclusion as the authors did not manipulate affective states in this study.

Introduction

“Due to both government legislation and consumer demands, the practice of housing of gestating gilts and sows as individuals in stalls has been banned in many developed countries, including the EU1, Australia, New Zealand, Canada, South Africa and ten states in the United States2,3”

This sentence could be read as stalls are forbidden in these countries. In practice animals can still be confined during the lactation period and sometimes for a few weeks during the gestation period. Can the authors add clarification here?

“The personality of an organism refers to the relatively consistent behavioral reactions of an organism across contexts and is most clearly expressed in novel and unpredictable situations17”

This sentence is not super clear + avoid repetitions (maybe replace organism by individuals or animals). Overall the authors do not clearly articulate how their measures can be used to reflect personality traits in sows.

“and the expression of pain21” Again there is more recent work that would better fit your article as variation in jbt responses has been linked to pain in calves (see Lecorps et al., 2020).

L70 to 74: I would push the authors to also consider the work being done in rats and dairy calves.

L75 the logic needs some work in this paragraph. The authors start by saying that there is not enough research being done on affective states (from this paragraph it feels like the authors say that research on affective states in pigs are limited to judgment bias studies). Then they say that there is not enough research being done on personality traits. However, they do not really stress out why we need to explore the relationship between response to jbt (hypothesized to reflect on mood) and personality traits. Why is this important?

One or two additional sentences would help the reader following the point made by the authors.

Overall, I think this is one of the points I’d like the authors to focus on. The conceptualization of the study and rationale for the question asked is not very clear in the introduction. The paper would be much stronger if these points were improved.

L95 will impact response to judgment bias tests

Methods

L158 a dot is missing

Personality tests used seem different from the rest of the literature. Can the authors comment on that? What is typically done? Why not mimicking what Asher et al 2016 (one of the key reference for this work) did?

I like the fact that the procedures intended to use routine procedures happening to piglets/sows in commercial settings, but I think this choice needs to be justified and explained.

L171 An illustration of my doubts regarding the first paragraph of the introduction. The readers may be led to think that the ban of stalls/crates means animals are no longer kept in confinement during their life, which is not the case. However, this still happens at various times.

L205 “sow moves easily with little inducement. She is comfortable on all her feet” I would link these two sentences, otherwise it is a little confusing. Same for the following sentences. Another option would be to have a table for this.

“Sows were considered trained if they approached the positive stimulus within 20 s for 100% of the positive trials and if they did not approach the negative stimulus at least 50 s for at least 80% of the negative trials.”

Can the authors explain why they chose these criteria? Do they come from previous studies?

L254 Why only one ambiguous trial? This is very unusual as most JBT studies would use at least multiple exposure to the same ambiguous cue or to multiple cues. How confident are the authors that this was enough? I feel this needs to be justified.

L259 It is not clear if the authors mean they used all variables collected in all the different tests (except JBT) or just “during pen introduction” (i.e. social interactions).

L268 please state here how many sows did not meet the learning criteria?

L282 “with latency being defined as the time to approach the specific stimulus cue” The authors just said they used an index so which variable was used for statistical analyses is not clear (i.e., raw latencies or JBI?).

L305 you mention > 0.6 in the method section.

Results

There are several points in the results and presentation of the results that are confusing.

First, what is the point of calculating an index if not to use it? I honestly prefer the raw latencies that are much easier to comprehend. If used, the authors need to explore whether responses to positive and negative cues affected response to the ambiguous cue (which seems to have been done later on in the analysis?).

Second, in Figure 3A. I do not understand why the authors categorized sows according to their response to the ambiguous cue here. It makes the reader thinking that the analysis is categorical when it is not. Typically, studies using JBT would display this type of graph to show that animals (at the population level) responded to the ambiguous cue as expected (i.e. in average responses are situated between the two learnt stimuli). As is, the figure makes us think that the authors removed the four animals that displayed intermediate responses from the analyses. Please modify this figure either by presenting means for all three locations or individual responses or a combination of both.

“Both positive and negative bias animals approach the positive and negative conditioned stimulus with similar latencies independent of their judgement bias.”

So what’s the point of the index then? In addition, I really don’t like the expressions “positive bias animals” and “negative bias animals” Why not talking about “optimistic” and “pessimistic” animals like the rest of the literature? Alternatively, you could also talk about animals displaying positive and negative expectations.

“It appears that these animals failed to exhibit such a preference as the others.”

I am not sure I understand what the authors mean here. A preference for what?

L365 Can the authors better explain why they did not test the effect of probe location on latencies independently from the other variables?

L365 Again, here it feels like the authors are using the raw variables instead of the individual coordinates obtained from the three dimensions of the PCA. What’s the point of the PCA then if not to summarize information?

As far as I understand correctly, the model should look like: latency = probe location * aggressiveness + probe location * submissiveness + probe location * activity.

If yes, then why do we have 3 variables that were used to assess the trait “aggressiveness” tested separately?

Also, it seems that the authors are using different terms for the same variables and it makes it difficult to follow. So maybe I am just getting confused here.

Figure 4 could also be improved. R2 are missing and usually regressions are represented using squared frame (same width and height) so that regression lines are easier to interpret.

Looking at the relationship between aggressiveness and JB, I wonder if the authors considered a quadratic model? As displayed it feels like more aggressive sows were more optimistic but less aggressive ones were either optimistic or pessimistic. Thus, not a true linear relationship that is consistent with what Asher found if my memories are correct. This point needs to be discussed.

Discussion

“Both positive and negative judgement bias”. It feels odd to phrase it this way (see previous comment). I would suggest that the authors change for individual differences in judgment bias (here and elsewhere in the paper) for two reasons: 1) we don’t know when a positive and a negative bias starts. 2) it puts a lot of emphasis on the extremes and tend to ignore the four sows that expressed true intermediate responses. How should they be considered? as neutral judgment bias? Using “inter-individual differences” will solve these issues and be more consistent with the literature.

“Perhaps surprisingly, we did not find that animals with more severe physical indicators of poor welfare, such as reduced body condition or increased lesion severity, displayed more negative judgement biases”

Well is this really surpsising considering that the authors just explained that lameness and BCS were not really variable in their cohort?

These two elements should be removed from all conclusion sentences as the authors don’t know whether more lame or more thin animals would express negative judgment bias. The authors could add a sentence stating that “variation in BCS and Lameness was too small for meaningful analyses” and only comment on the severity of the lesions.

L428 is this a continuum though? The authors’ data suggest that submissiveness and aggressiveness are independent (on two axes of the PCA). This is consistent with recent research on dominance in sows, which should be incorporated here.

“Taken together, these results suggest that the judgement bias paradigm used in the present study more likely was evaluating a sow’s response to risk as it relates to consistent individual differences in behavior (i.e., personality) rather than exclusively evaluating the subjective mood state (i.e., mental welfare) of the sow”

How do you know? Maybe the “personality differences” you assessed reflected or were affected by the overall mood states of the animals.

I’d like the authors to integrate the growing research that explore judgment biases as states and traits. See research being done in dairy calves for more information on this aspect. I think this will help the authors phrase their results.

“locomotion-based” well this type of test is typically referred to as a spatial learning task.

L436 – 458 I agree with the authors. Similar issues were found in laboratory animals when the relationship between stereotypic behaviours and response to jbt were assessed.

L460 “(rats, dogs, parrots)” but you cite studies in pigs, dogs and primates.

L461 Which personality traits though? To my knowledge none of these studies looked at aggressiveness per se.

In addition, there is much more in the literature than these three studies and it may help the authors to look more deeply at what has been found in species such as rats, dairy calves, dolphins and many others. Some of these studies specifically looked at dominance, social behaviours, fearfulness etc. All these elements could help the authors better interpreting their results.

I also want to attract the attention of the authors that some work being done in dogs and primates are harder to interpret given that questionnaires are sometimes used in dogs and low sample size are frequent in primates. These methodological differences may render comparisons difficult.

L464 to 473. It is a little difficult to follow the authors here. Please try to rewrite and clarify.

L476 why not, but do the authors have evidence that injuries lead to small differences in subjective state?

L480 yes! Thus, I am sorry to say that calling the behavioural assessment done in these sows “personality traits” is a bit misleading (see previous comments), unless the authors can better justify?

L511 a very recent study in mice support the point being made here (see Jardim et al., 2021)

L516 yes! You will see that following studies in calves from the same group explored the logic you described at the end of the discussion.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

29 Oct 2021

Revision note

PONE-D-21-16018

Judgement bias of group housed gestating sows predicted by personality traits, but not physical measures of welfare

Editor:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming.

Response: PLOS ONE style requirements were addressed.

2, We note that the grant information you provided in the ‘Funding Information’ and ‘Financial Disclosure’ sections do not match. When you resubmit, please ensure that you provide the correct grant numbers for the awards you received for your study in the ‘Funding Information’ section.

Response: This has been addressed during re-submission.

3. Thank you for stating the following in the Acknowledgments Section of your manuscript:

“The authors would like to thank the summer research interns of the Swine Research and Teaching Center for their assistance in data collection. This work was supported in part by Kraft Foods Oscar Mayer, Pig Improvement Company, Pennsylvania Pork Producers Council, Pennsylvania Soybean Board, Pennsylvania Department of Agriculture, and the American Society for the Prevention of Cruelty to Animals”

We note that you have provided additional information within the Acknowledgements Section that is not currently declared in your Funding Statement. Please note that funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form. Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

“This work was supported in part by Kraft Foods Oscar Mayer, Pig Improvement Company, Pennsylvania Pork Producers Council, Pennsylvania Soybean Board, Pennsylvania Department of Agriculture, and the American Society for the Prevention of Cruelty to Animals. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript”

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

Response: We have removed all references to funding support from the acknowledgments and included this information in the funding statement.

Reviewer #1: This is an important subject, to challenge the recent tendency to over-interpret results from cognitive bias paradigms as de facto evidence of good or bad welfare. It would be nice to see consideration of alternative explanations for cognitive bias results become the norm.

Reviewer #2: This paper aims at exploring what factors influence performance in judgement bias testing in group-housed sows with an emphasis on personality traits and measures of physical aggressions. This is an important area of research given the need for a better understanding of affective states of farmed animals and how it can successfully relate to other welfare measures. I think the study has merits, but I feel it could really be improved. I spent quite some time making both general and detailed comments that I hope will help the authors.

General comments

• The literature considered in this manuscript is rather incomplete as work on other species (dairy calves and rats) has not been integrated. I would strongly recommend including what has been done in other species given that only a few studies have explored individual variation in response to judgment bias. For now, the authors do not really refer to studies exploring this link but rather mention their existence. I think the integration of these studies will provide more context and should help the authors better communicate their results, especially in terms of how moods can be considered both as reflecting emotional states and/or personality differences.

Response: Previous studies which examine personality, coping style, or behavioral syndrome in a variety of specie, as it relates to risk-taking, appraisal or judgment bias, have been incorporated throughout the introduction and discussion.

• The rationale behind the research question could be improved. For now, the reader is left with limited information on why the authors did this study and why it is important.

Response: The last sentence in the second to last paragraph has been revised to clarify why this study was conducted with this particular population of swine.

• I’d like the authors to discuss more some limitations of their study. For instance, the authors used only one ambiguous cue presented only once, which is at odds with the literature. The authors need to explain why they made that choice and more effectively explain how this could have affected their results (i.e., animals may be more incline to react to ambiguous cues as novel element in their environment on the first exposure vs. on repeated exposures).

Response: The reasons for only one ambiguous cue (animal fatigue, chronic animal hunger) have been incorporated into the methods. We have addressed the limitation of one ambiguous trial in the discussion with greater details; including ways to improve upon our study design in the future.

• Many scientists would argue that personality traits need to be assessed over time and contexts. In addition, tests were not all standardized (e.g., the social interaction test was more a regrouping experiment), which is a limitation of the study. I would recommend to keep the concept of personality for the discussion but I may be convinced otherwise if the authors clearly stress the consistency of their outcome variables.

Response: We agree with this critique. We have replace the word ‘personality traits’ with ‘behavioral traits’ throughout the manuscript, but maintain the theme of individual differences in behavior consistent over time were found with the dataset; as we have established this difference using the same study cohort in another peer reviewed study (Horback and Parsons, 2018, Behavioral Processes).

• What is the point of the JBI? I would remove this part as the authors do not seem to use it for further analyses.

Response: The Judgment Bias Index was created to control for differences in the approach speed between sows (i.e., some animals simply always moved slower than others). However, in this study there was not that much difference in locomotion rates between sows and hence why as the reviewers point out, this index is not used for further analysis or interpretation. As such we have removed all information about the JBI from this manuscript.

Abstract

L21 states?

Response: The ‘s’ has been added.

L22 I feel the authors should give more information to the reader so that it is easier to grasp how JBTs work. I would add elements explaining pessimistic and optimistic responses.

Response: Sentences have been added to explain the optimistic/pessimistic interpretation of animal behavior in JBTs.

L26 Maybe better to split this sentence?

Response: This sentence has been split into two.

L36 I would tone down the conclusion as the authors did not manipulate affective states in this study.

Response: This final sentence to the abstract has been revised to reduce the strengths of conclusions to be taken from our data.

Introduction

“Due to both government legislation and consumer demands, the practice of housing of gestating gilts and sows as individuals in stalls has been banned in many developed countries, including the EU1, Australia, New Zealand, Canada, South Africa and ten states in the United States2,3”

This sentence could be read as stalls are forbidden in these countries. In practice animals can still be confined during the lactation period and sometimes for a few weeks during the gestation period. Can the authors add clarification here?

Response: This sentence has been revised.

“The personality of an organism refers to the relatively consistent behavioral reactions of an organism across contexts and is most clearly expressed in novel and unpredictable situations17”

This sentence is not super clear + avoid repetitions (maybe replace organism by individuals or animals). Overall the authors do not clearly articulate how their measures can be used to reflect personality traits in sows.

Response: We have included a more thorough definition of personality and included a segue sentence to describe how tests to evaluate individual differences in behavior should be novel or unpredictable to reveal high-arousal responses.

“and the expression of pain21” Again there is more recent work that would better fit your article as variation in jbt responses has been linked to pain in calves (see Lecorps et al., 2020).

Response: This source has been incorporated to the introduction.

L70 to 74: I would push the authors to also consider the work being done in rats and dairy calves.

Response: Given that there is work done relating personality to performance in JBTs for many other species (dolphins, mice, rats, dairy cows, etc.), we have decided to expand on this relationship in the discussion, rather than elongate the introduction. This sentence does state that the following studies are examples of this relationship.

L75 the logic needs some work in this paragraph. The authors start by saying that there is not enough research being done on affective states (from this paragraph it feels like the authors say that research on affective states in pigs are limited to judgment bias studies). Then they say that there is not enough research being done on personality traits. However, they do not really stress out why we need to explore the relationship between response to jbt (hypothesized to reflect on mood) and personality traits. Why is this important? One or two additional sentences would help the reader following the point made by the authors. Overall, I think this is one of the points I’d like the authors to focus on. The conceptualization of the study and rationale for the question asked is not very clear in the introduction. The paper would be much stronger if these points were improved.

Response: This paragraph has been revised to include research on assessing swine affective states other than JBTs, as well as to highlight the lack of information on breeding sow personality (instead of piglet personality). Sentences were added to clarify the need for more information on if/how sow personality influences performance in JBT (as little is known about this breeding population).

L95 will impact response to judgment bias tests

Response: This change has been made.

Methods

L158 a dot is missing

Personality tests used seem different from the rest of the literature. Can the authors comment on that? What is typically done? Why not mimicking what Asher et al 2016 (one of the key reference for this work) did? I like the fact that the procedures intended to use routine procedures happening to piglets/sows in commercial settings, but I think this choice needs to be justified and explained.

Response: A sentence has been added to explicitly state and defend the use of these tests which can be incorporated into routine animal husbandry, for a more generalizable and practical use. It should also be noted that the study population is different between our study and Asher et al. In this manuscript, we follow animals from 5 weeks of age to just beyond conception of their second litter (a 50 week or more time span) whereas Asher et al focused on grower pigs that varied in age over a 6 week period from 4 to 10 weeks of age. The personality test that are used here are previously described in the following publications (Horback and Parsons, 2016 & 2018).

L171 An illustration of my doubts regarding the first paragraph of the introduction. The readers may be led to think that the ban of stalls/crates means animals are no longer kept in confinement during their life, which is not the case. However, this still happens at various times.

Response: The first paragraph of the introduction has been modified to reduce the insinuation that sows are never in stalls or crates.

L205 “sow moves easily with little inducement. She is comfortable on all her feet” I would link these two sentences, otherwise it is a little confusing. Same for the following sentences. Another option would be to have a table for this.

Response: The description of the lameness scale was revised to have one sentence per score to reduce confusion.

“Sows were considered trained if they approached the positive stimulus within 20 s for 100% of the positive trials and if they did not approach the negative stimulus at least 50 s for at least 80% of the negative trials.”

Can the authors explain why they chose these criteria? Do they come from previous studies?

Response: These criteria are based on average walking speed of the sows and a clear break in the performance of the sows looking at the data (explained in Horback & Parsons, 2018, Behavioral Processes). This explanation has been added.

L254 Why only one ambiguous trial? This is very unusual as most JBT studies would use at least multiple exposure to the same ambiguous cue or to multiple cues. How confident are the authors that this was enough? I feel this needs to be justified.

Response: We understand this concern as all other spatial judgement bias tasks include probe cues in the assessment. We have included in the methods that our pregnant sows were not able to be placed in multiple trials, and thus additional probe cues, as they became physically exhausted from standing and walking after 30 minutes. Not described in the manuscript, but for the reviewers’ knowledge, we have developed a non-strenuous activity (joystick movement from stationary/sitting animal) to allow for probe cues to be used in future judgment bias tasks with breeding sows.

L259 It is not clear if the authors mean they used all variables collected in all the different tests (except JBT) or just “during pen introduction” (i.e. social interactions).

Response: This sentence has ben revised to clarify that it is only the behaviors during introduction to pen as adults that were placed in PCA.

L268 please state here how many sows did not meet the learning criteria?

Response: This information has been added.

L282 “with latency being defined as the time to approach the specific stimulus cue” The authors just said they used an index so which variable was used for statistical analyses is not clear (i.e., raw latencies or JBI?).

Response: The Judgment Bias Index was created to standardize the different approach speed between sows (i.e., some animals simply always moved slower than others). But as the reviewers point out, this index is not used for further analysis or interpretation, and therefore, we have removed all information about the JBI from this manuscript.

L305 you mention > 0.6 in the method section.

Response: The cut-off for ‘strong loading’ will remain 0.6. This sentence has been revised.

Results

There are several points in the results and presentation of the results that are confusing. First, what is the point of calculating an index if not to use it? I honestly prefer the raw latencies that are much easier to comprehend. If used, the authors need to explore whether responses to positive and negative cues affected response to the ambiguous cue (which seems to have been done later on in the analysis?).

Response: The Judgment Bias Index was created to standardize the different approach speed between sows (i.e., some animals simply always moved slower than others). But as the reviewers point out, this index is not used for further analysis or interpretation, and therefore, we have removed all information about the JBI from this manuscript.

Second, in Figure 3A. I do not understand why the authors categorized sows according to their response to the ambiguous cue here. It makes the reader thinking that the analysis is categorical when it is not. Typically, studies using JBT would display this type of graph to show that animals (at the population level) responded to the ambiguous cue as expected (i.e. in average responses are situated between the two learnt stimuli). As is, the figure makes us think that the authors removed the four animals that displayed intermediate responses from the analyses. Please modify this figure either by presenting means for all three locations or individual responses or a combination of both.

Response: We agree with the reviewer that there is a better way to display this data. The JBI data has been removed from the manuscript, and the original Figure 3A an 3B has also been removed. We have included a new Figure 3, which displays the data points for individual animal approaching the three difference cue locations as well as a mean value for each animal group (optimistic, pessimistic, and undetermined bias).

“Both positive and negative bias animals approach the positive and negative conditioned stimulus with similar latencies independent of their judgement bias.”

So what’s the point of the index then? In addition, I really don’t like the expressions “positive bias animals” and “negative bias animals” Why not talking about “optimistic” and “pessimistic” animals like the rest of the literature? Alternatively, you could also talk about animals displaying positive and negative expectations.

Response: We have removed all references and uses of the JBI from the manuscript, and are focusing all analyses and interpretations on the raw latencies. We have also replaced all terms of ‘positive or negative bias” with the terms “optimistic or pessimistic bias” to be consistent with the literature.

“It appears that these animals failed to exhibit such a preference as the others.”

I am not sure I understand what the authors mean here. A preference for what?

Response: This sentence has been revised to state, “The response of four sows was intermediate between the more optimistic or more pessimistic responses of others sows, and therefore, were labeled as having an undetermined bias”.

L365 Can the authors better explain why they did not test the effect of probe location on latencies independently from the other variables?

Response: Additional probes were not used in the design given the physical limitations of the sows. We have included in the methods that our pregnant sows were not able to be placed in multiple trials, and thus additional probe cues, as they became physically exhausted from standing and walking after 30 minutes. In addition, we argue that there is a significant interaction between probe location and latency of approach can be seen in Figure 3.

L365 Again, here it feels like the authors are using the raw variables instead of the individual coordinates obtained from the three dimensions of the PCA. What’s the point of the PCA then if not to summarize information? As far as I understand correctly, the model should look like: latency = probe location * aggressiveness + probe location * submissiveness + probe location * activity. If yes, then why do we have 3 variables that were used to assess the trait “aggressiveness” tested separately? Also, it seems that the authors are using different terms for the same variables and it makes it difficult to follow. So maybe I am just getting confused here.

Response: We have revised this section to clarify that 10 behaviors were compressed using PCA to create 3 composite traits. This way, only one trait score for “aggressiveness” among the sows was compared to approach latencies (instead of 6 separate behavioral comparisons to approach latencies [i.e., bite, chase, displace, etc.]). The behavior of a sow during the ‘litter restraint test’ was removed from the PCA, as it is evaluated as a separate behavioral measure. We have added to this section, “There were no significant effects of the submissive or active/exploratory trait scores on the latency to approach positive, ambiguous or negative stimulus.”

Figure 4 could also be improved. R2 are missing and usually regressions are represented using squared frame (same width and height) so that regression lines are easier to interpret.

Response: The graphs of Fig 4 have been revised, and now include r-squared values.

Looking at the relationship between aggressiveness and JB, I wonder if the authors considered a quadratic model? As displayed it feels like more aggressive sows were more optimistic but less aggressive ones were either optimistic or pessimistic. Thus, not a true linear relationship that is consistent with what Asher found if my memories are correct. This point needs to be discussed.

Response: Asher et al reported interactions between cue location, personality trait and housing type, but all were described with a simple linear mixed model. We appreciate the reviewer’s point about the variability/noisiness of the latency response as the aggressive/dominant trait score decreases. We believe that given this variability we would need significantly more observations to successfully fit a more complicated model such a quadratic and thus opted to use a simple linear mixed model as others like Asher et al have done before.

Discussion

“Both positive and negative judgement bias”. It feels odd to phrase it this way (see previous comment). I would suggest that the authors change for individual differences in judgment bias (here and elsewhere in the paper) for two reasons: 1) we don’t know when a positive and a negative bias starts. 2) it puts a lot of emphasis on the extremes and tend to ignore the four sows that expressed true intermediate responses. How should they be considered? as neutral judgment bias? Using “inter-individual differences” will solve these issues and be more consistent with the literature.

Response: We have also replaced all terms of ‘positive or negative bias’ with the terms “optimistic or pessimistic bias” to be consistent with the literature. We have also included a statement about the ‘undetermined bias’ sows from our study in this section.

“Perhaps surprisingly, we did not find that animals with more severe physical indicators of poor welfare, such as reduced body condition or increased lesion severity, displayed more negative judgement biases”

Well is this really surprising considering that the authors just explained that lameness and BCS were not really variable in their cohort? These two elements should be removed from all conclusion sentences as the authors don’t know whether more lame or more thin animals would express negative judgment bias. The authors could add a sentence stating that “variation in BCS and Lameness was too small for meaningful analyses” and only comment on the severity of the lesions.

Response: We have removed the emphasis on body condition and lameness as indicators of welfare condition, and included the suggested statement of the lack variation of these measures.

L428 is this a continuum though? The authors’ data suggest that submissiveness and aggressiveness are independent (on two axes of the PCA). This is consistent with recent research on dominance in sows, which should be incorporated here.

Response: The word ‘continuum’ has been removed from this sentence, and previous studies which found that lesion scores among sows is not a direct proxy for aggression measurement have been cited.

“Taken together, these results suggest that the judgement bias paradigm used in the present study more likely was evaluating a sow’s response to risk as it relates to consistent individual differences in behavior (i.e., personality) rather than exclusively evaluating the subjective mood state (i.e., mental welfare) of the sow” How do you know? Maybe the “personality differences” you assessed reflected or were affected by the overall mood states of the animals.

I’d like the authors to integrate the growing research that explore judgment biases as states and traits. See research being done in dairy calves for more information on this aspect. I think this will help the authors phrase their results.

Response: We have included more literature which examines judgment bias and personality in other species throughout the discussion. In addition, we have included a short review of the theories behind individual differences in appraisal of stimuli as it relates to state (mood) and trait (affective style) (i.e., Faustino et al., 2015).

“locomotion-based” well this type of test is typically referred to as a spatial learning task.

Response: These phrase ‘locomotion based’ has been replace with “spatial learning task”.

L436 – 458 I agree with the authors. Similar issues were found in laboratory animals when the relationship between stereotypic behaviours and response to jbt were assessed.

Response: We appreciate the reviewer’s recognition of the relevance for this section.

L460 “(rats, dogs, parrots)” but you cite studies in pigs, dogs and primates.

Response: This sentence has been removed.

L461 Which personality traits though? To my knowledge none of these studies looked at aggressiveness per se. In addition, there is much more in the literature than these three studies and it may help the authors to look more deeply at what has been found in species such as rats, dairy calves, dolphins and many others. Some of these studies specifically looked at dominance, social behaviours, fearfulness etc. All these elements could help the authors better interpreting their results.

Response: This sentence has been removed.

I also want to attract the attention of the authors that some work being done in dogs and primates are harder to interpret given that questionnaires are sometimes used in dogs and low sample size are frequent in primates. These methodological differences may render comparisons difficult.

Response: We agree with the reviewer, that personality determined from human ratings, rather than ethological coding, can be difficult to interpret (or generalize to other animal personality studies). Therefore, we have reduced the reference to such studies and focused on referring to studies with similar methodology as our present study.

L464 to 473. It is a little difficult to follow the authors here. Please try to rewrite and clarify.

Response: This paragraph has been rewritten to clarify our point that there are themes within the behavioral syndrome, coping style, and personality literature which link the traits related to survival (i.e. large risk taking for big reward).

L476 why not, but do the authors have evidence that injuries lead to small differences in subjective state?

Response: We have removed this section of the sentence to prevent the insinuation that we assume small injuries would lead to differences in mood state.

L480 yes! Thus, I am sorry to say that calling the behavioural assessment done in these sows “personality traits” is a bit misleading (see previous comments), unless the authors can better justify?

Response: We have replaced the word ‘personality traits’ with ‘behavioral traits’ throughout the manuscript, but maintain the theme of individual differences in behavior consistent over time were found with the dataset; as we have established this difference using the same study cohort in another peer reviewed study (Horback and Parsons, 2018, Behavioral Processes).

L511 a very recent study in mice support the point being made here (see Jardim et al., 2021)

Response: The Jardim et al (2021) paper has been incorporated into this sentence.

L516 yes! You will see that following studies in calves from the same group explored the logic you described at the end of the discussion.

Response: We have incorporated additional studies on animal personality, judgment bias, and state vs trait of appraisal analysis throughout the introduction and discussion based on the recommendations of the reviewer.

Submitted filename: PLOS ONE revision note CLEAN.docx

Click here for additional data file.

22 Nov 2021

31 Dec 2021

Revision note

12/28/2021

PONE-D-21-16018

Judgement bias of group housed gestating sows predicted by personality traits, but not physical measures of welfare

Reviewer #2: First, I want to thank the authors for a great revision of their paper. I think it has greatly improved. I also think the study contributes to highlight a growing problem in the field: what do we do with animal welfare markers that do not correlate? I would like the authors to consider putting more emphasis on this point rather than concluding that JBTs provide limited information. We actually do not know which marker is the most limited. It is also interesting to put forward that when JBT are used in absence of an emotional manipulation, it typically provides information about individual characteristics. I made a few more comments that I’d like the authors to consider (see below).

Response: We thank the reviewer for providing very helpful feedback to improve this manuscript. We agree that different welfare metrics will have different limitations and is a challenge in elucidating the overall affective state of the animal. In the revised version of this manuscript, we suggest that individual behavioral characteristics could create differential emotional responses that are captured in the JBT. Perhaps an animal’s boldness or risk attributes determine how they perceive a complicated social environment like pen gestation and as such influences their emotional response. We have amended the introduction and discussion to place more emphasis on the role of cognitive bias testing; that is, how do we interpret response latency given the possibility that different factors (eg behavioral traits and lesion) may be differentially contributing to the animal’s affective state.

Abstract:

The authors make a very strong claim here (i.e., that JBT may not “be the most appropriate method for evaluating affective state”), but I am not convinced. This argument is based on 1) high variability in responses to JBT in enriched animals, 2) a positive relationship with aggressiveness and 3) a lack of relationship with other so-called welfare markers (e.g., mostly the number of lesions after regrouping).

First, yes, there is high variability, but this does not mean swine kept in barren conditions would not score worse. Second, yes it correlates with aggressiveness, which makes sense given that aggressiveness often correlates with proactive coping styles, impulsivity and optimism. I’m not sure if the authors saw this recent review (Lecorps et al., 2021. Negative expectations and vulnerability to stressors in animals) but it seems very relevant for this study! Third, are these physical indicators truly reflecting what animals feel or their welfare state? Are only unhealthy humans feeling terrible?

Response: Not sure, what the reviewer means in terms of sows in barren conditions scoring worse. Our goal is not the response of the group but that fact that several individual animals essentially interpret the ambiguous stimulus as the negative cue. I do not know how these animals could score worse even if housed in barren conditions. However, this does speaks towards the notion that personality attributes could be creating the emotional experience captured in this JBT. We also agree that physical compromise could, but necessarily, result in negative affective states and tried to better explain the observed disconnect between lesions scores and JBT outcomes. The abstract has been modified to capture these nuances.

I think the authors should discuss more this aspect because it is a very important one and they are not the only ones that failed to show correlations between psychological and physical markers of well-being (e.g., Lecorps et al. 2021. Regrouping induces anhedonia-like responses in dairy heifers; in that study no correlations between anhedonia-like behaviours and the number of agonistic interactions received during regrouping were found). Importantly, I think this does not necessarily mean that the methodologies do not work well but rather that our conception of welfare is somewhat challenged. I would like the authors to include this aspect (i.e., the difficulty to correlate different welfare parameters) in their conclusions. As is, the claims are not largely supported by the authors’ dataset.

Response: We appreciate the reviewer’s request to make this point more explicit in our paper. We have added the following lines to our discussion to highlight that this disconnect between physical and psychological welfare assessments (lines 441-449): “Current welfare assessments for breeding sows focus on physical ailments (e.g., lameness, body condition and skin lesions), as well as being the target of aggressive behaviors (e.g., bites, scratches, displacement). However, given that recent work in dairy heifer welfare has reported a lack of correlation between behavioral indicators of anhedonia and the number of agonistic interactions received during regrouping (Lecorps et al., 2021), there is a need to carefully critique our assumptions of how welfare is experienced by the animal. If we are using judgement bias testing as an indicator of subjective mental welfare of animals, we also need to better understand how, and if, the degree and duration of physical comprise contribute to an animal’s psychological welfare state.”

L54 There are several reviews addressing cumulative affective experiences and mood-related disorders in farm animals (Poirier et al., 2019, Bateson and Poirier, 2019, Lecorps et al., 2021, McLellan and Mason, 2021), maybe worth mentioning at least one of these so that the interested reader can learn more?

Response: We have revised this sentence to clarify that little attention is currently being given to breeding sow affective state (and not farm animals in general), but we have added a citation to this sentence to allow the reader to know of the growth of knowledge on the subject of affective states in farm animals.

L69 to 73 I haven’t read this particular contribution of Stamps but I find some of the arguments quite surprising. For instance, what do “early in life” or “biologically based” mean? Are all traits more clearly expressed in “novel and unpredictable situations” or just the ones we care to measure in animals? It makes perfect sense if we assess neophobia, boldness or pessimism but does it make sense for sociability and aggressiveness? In my opinion, these are risky claims, but the authors are of course free to keep this section if they feel it helps with the narrative.

Response: This summary of personality has been revised to reduce any claims of genetically-linked traits, or strict consistency over an organism’s lifetime. This sentence now reads, “Personality can be described as characteristics of behavioral reactions such as intensity, speed, response threshold, latency and recovery time, which are most clearly expressed in novel and unpredictable situations, and are relatively stable throughout an individual’s development.”

L97 “Given that there is limited information on the use and validity of judgement bias testing to evaluate affective states in sexually-mature sows, there is a need to investigate possible influences on performance in this cognitive task. As greater emphasis is being placed on the welfare of commercial sows, because they live the longest out of all production swine and thus have the greatest capacity to incur chronic suffering if their welfare is compromised, investigating new tools for welfare assessment is crucial.”

The second sentence could be deleted as it basically repeats what is said earlier (L85). I’m still not sure to understand the aims as currently formulated.

Response: The second sentence has been removed to prevent redundancy. A sentence clarifying the aim of this study has been added in its place.

L102 to 108 maybe this part belongs to the method section?

Response: These sentences have been removed as they summarize the methods of this study.

L112 “if judgement bias reflects the affective state of an animal (i.e., biological health and mental state), then it follows that physical measurements of welfare (i.e., skin lesions, body condition) should also impact performance in judgement bias task”

As argued before this may not be necessarily true. However this is a good working hypothesis.

Response: We have kept this portion of the hypothesis in place, as we agree with the reviewer that physical health may not be the best predictor of subjective mental state; this notion is assumed in some animal emotion/welfare literature.

L278 For clarity sake I would start this sentence by “Additional probe cues”.

Response: This change has been made.

L358 to 363 this basically repeats the previous sentences, delete?

Response: These sentences have been deleted.

Figure 3 seems to have changed based on caption changes but I could not access it (neither any other figures).

Response: We do not know why the reviewer could not access the new Figure 3.

Figure 4 has also been changed according to the authors’ response, but I could not see the changes.

Response: We do not know why the reviewer could not access the new Figure 4.

L446 affective or coping style?

Response: This change has been made.

L454 “we need to reconcile inconsistencies when comparing the physical and assumed psychological welfare state.” I’d push the authors to consider that these may not be inconsistencies. Health being a poor predictor of psychological well-being in humans, it may well be the same in animals. However, I agree with the authors that lameness should, in theory, correlate with judgment bias given its severe impact on the animals’ quality of life.

Response: We agree with the reviewer on this suggestion, and have added a sentence the following sentence, “If we are using judgement bias testing as an indicator of subjective mental welfare of animals, we also need to better understand how, and if, the degree and duration of physical comprise contribute to an animal’s psychological welfare state.”

L479 there is a problem with this sentence (missing word?).

Response: This sentence has been removed.

L519 “it can be argued”. There are several typos along the manuscript that the authors will want to address before publication.

Response: The word “be” has been added.

Submitted filename: PLOS ONE revision note #3 CLEAN.docx

Click here for additional data file.

8 Feb 2022

Judgement bias of group housed gestating sows predicted by behavioral traits, but not physical measures of welfare

PONE-D-21-16018R2

Dear Dr. Horback,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Miquel Llorente, PhD

Academic Editor

PLOS ONE

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: No

11 Feb 2022

PONE-D-21-16018R2

Judgement bias of group housed gestating sows predicted by behavioral traits, but not physical measures of welfare

Dear Dr. Horback:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Miquel Llorente

Academic Editor

PLOS ONE