Applied Animal Behaviour Science 259 (2023) 105836 Contents lists available at ScienceDirect Applied Animal Behaviour Science journal homepage: www.elsevier.com/locate/applanim Play behaviour positively relates to weight gain, feeding behaviour and drinking behaviour in weaner pigs (Sus scrofa) Guilherme A. Franchi a, *, Mona L.V. Larsen a, b, Ida H. Kristoffersen a, Jeanet F.M. Winters a, Lene Juul Pedersen a, Margit Bak Jensen a a Aarhus University, Department of Animal and Veterinary Sciences, Blichers Allé 20, 8830 Tjele, Denmark b M3-BIORES: Measure, Model and Manage Bioresponses, Division Animal and Human Health Engineering, Department of Biosystems, KU Leuven, Kasteelpark Arenberg 30, 3001 Heverlee, Belgium A R T I C L E I N F O A B S T R A C T Keywords: Engagement in play behaviour has been associated with the presence of positive affective states and, thus, Positive animal welfare proposed to be an indicator of positive animal welfare. However, the interpretation of play in animals remains Locomotor-rotational play challenging due to the complexity of motivating factors. Accordingly, we aimed to clarify whether Yorkshire × Social play Landrace weaner pigs would engage more in play behaviour the more well-nourished they were by examining Early weaning the effects of weight gain, feeding behaviour, and drinking behaviour on two types of play behaviour [locomotor- Growth Hunger rotational play (LOC) and social play (SOC)]. In total, 24 litters [pigs/litter: (mean ± SD) 13 ± 2] raised under conventional husbandry conditions were included in this study. Each pig was manually weighed within 24 h of birth and on days − 7, 0, 1, 2 relative to the weaning day (day 0) at approximately 26 days of age. All behavioural measures were registered via video at individual level. Visits to feeder and drinker were registered from 07:00 h to 21:59 h on days − 1 and 1 using 2-min interval instantaneous sampling. The proportion of visits to each resource was calculated by dividing the number of scans visiting the resource by the total number of daily scans. The latencies to visit the feeder and drinker within the first 24 h post-weaning were continuously recorded. Both LOC and SOC were registered between 14:00 h and 22:00 h on days − 1, 1 and 2. Both before and after weaning, heavier pigs spent more time performing LOC. Before weaning, heavier pigs spent more time per­ forming SOC. Proportion of visits to the feeder positively related to LOC after weaning. On the day before weaning, the proportion of visits to the drinker positively related to LOC. No clear relationships between the latency to feed and drink after weaning and play behaviour were found. Our study supports the hypothesis that motivation to play is higher when animals are in more stable conditions, e.g., well-nourished, and healthier than under suboptimal conditions. However, the fact that the nutritional measures did not similarly affect LOC and SOC suggests that these two types of play behaviour may be differently affected by the weaning context and questions whether they have the same underlying motivation. This study represents a step toward the validation of play as a positive animal welfare indicator. 1. Introduction or psychological challenges (Burghardt, 2005; Burgdorf and Panksepp, 2007). In fact, play behaviour is considered an energetically and The performance of play behaviour occurs during the physical, cognitively demanding activity that is displayed in a relatively safe and physiological, and psychological development of mammals (Siviy, 1998; stable environment (Špinka et al., 2001). Furthermore, the performance Špinka et al., 2001). Prior work reported that juvenile mammals typi­ of play behaviour is thought to be reinforcing and associated with pos­ cally perform this behaviour when their primary needs, such as food and itive affective states (Burghardt, 2005). With the increasing concern that safety, have been met (Newberry et al., 1988; Špinka et al., 2001; Held the welfare of animals should be safeguarded by not only avoiding the and Špinka, 2011). Conversely, a low or no engagement in play experience of negative affective states, but also promoting the experi­ behaviour in juveniles suggests the presence of physical, physiological, ence of positive ones, play behaviour has been suggested as a potential * Corresponding author. E-mail address: amorimfranchi@anivet.au.dk (G.A. Franchi). https://doi.org/10.1016/j.applanim.2023.105836 Received 25 October 2022; Received in revised form 11 January 2023; Accepted 16 January 2023 Available online 17 January 2023 0168-1591/© 2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). G.A. Franchi et al. Applied Animal Behaviour Science 259 (2023) 105836 positive animal welfare indicator (Boissy et al., 2007; Yeates and Main, TN-70 Yorkshire × Landrace (TN) gilts were included in one batch. 2008). Following farrowing, the twelve sows returned in a second batch as Undernourishment and hunger are acknowledged to negatively second-parity sows. Both genetic hybrids were inseminated with semen impact animal welfare and to limit the performance of play behaviour of DanBred Duroc boars. A total of 306 pigs (50% castrated males, 50% (Špinka et al., 2001; Boissy et al., 2007). For instance, in dairy calves, females) from 24 litters (pigs/litter: (mean ± SD) 13 ± 2; birth weight: low milk allowance (Duve et al., 2012; Jensen et al., 2015) and pre­ 1.4 ± 0.3 kg (both DB and TN)) across both batches were studied. The mature weaning (Miguel-Pacheco et al., 2015; Rushen et al., 2016) farrowing unit had two sections, each containing eight 3.0 × 2.2 m induced a decrease in locomotor play duration. Similarly, earlier studies farrowing pens for loose-housed sows (FT30, Jyden A/S, Denmark; reported a reduction in social play duration in undernourished wild approximately 0.51 m2/pig). Each pen had 41% slatted-flooring and animals (e.g., Hanuman langurs (Presbytis entellus) – Sommer and 59% solid-concrete-flooring (heated) and was equipped with sloped Mendoza-Granados, 1995; primates – Loy, 1970; wild meerkats (Suricata walls to reduce the risk of pig crushing by the sow, a 0.9 m2 hayrack next suricatta) – Sharpe et al., 2002). On the contrary, food rationing to the covered pig creep area (0.9 m2; equipped with a 150 W infrared reportedly led to an increase in object play duration in domestic kittens heat lamp), a low 80 × 28 cm polyconcrete feeder (Jyden A/S, Denmark; (Felis cattus) (Bateson et al., 1990; Hall and Bradshaw, 1998) and social Fig. 1A) filled with sow lactation diet (Die Profil Lac, DLG, Denmark, play duration in juvenile rats (Rattus norvegicus) (Loranca et al., 1999). 15.6% crude protein) allowing pigs to eat together with the sow, and a In the kittens’ example, object play may be motivated similarly to 31 × 17 × 11 cm pig drinker (Aqua-Level system with hinged trough, predatory behaviour, which can be stimulated when food availability is Jyden A/S, Denmark; Fig. 1A) allowing pigs to drink with the sow. The limited. In the rats’ example, social play may be difficult to distinguish artificial light was on from 07:00 h to 21:59 h. Every morning, feeders from agonistic interactions. Furthermore, for both kittens and juvenile were emptied before first feeding and manure was removed. Straw racks rats, the authors speculated that those young animals interpreted the were filled with wheat straw daily and approximately 130 g of fine reduced food supply as a cue to accelerate their development and chopped wheat straw were provided on the solid floor daily. Approxi­ engagement in play behaviour while it was still feasible. These con­ mately 400 g of sawdust were provided in the pig creep area daily. At trasting findings illustrate the complexity of play behaviour and un­ farrowing, the room temperature was set at 21 ◦ C and gradually derline the need for studies assessing the validity of various types of play decreased to 19 ◦ C at weaning. Sows received analgesic (Metacam®) at behaviour as positive animal welfare indicators against other validated farrowing and all pigs were weighed, ear tagged and had their umbilical welfare measures (Martin and Caro, 1985; Špinka et al., 2001; cord cut following farrowing (if it ended before 15:00 h) or on the next Ahloy-Dallaire et al., 2018). morning (if farrowing ended after 15:00 h). The number of own pigs In a natural setting, the period the sow nurses her offspring gradually nursed by the sows was equalised to the number of functional teats. declines from approximately 10 days post-partum (pp) and by 4–5 Surplus pigs within sow hybrid were cross fostered within the first 3 days months of age nursing ends and pigs become fully habituated to solid pp to sows that had fewer pigs than functional teats. If no sows were feed (Jensen and Stangel, 1992). In contrast, pigs raised in modern available for cross fostering, surplus pigs were euthanised by blunt force production systems are abruptly weaned from their sow and shift from trauma. Non-viable pigs and pigs below 700 g were euthanised first milk to solid feed at approximately 3–5 weeks of age (Špinka, 2017). followed by randomly selected pigs. On day − 1 relative to weaning, the Hence, weaning is a management practice known to inflict nutritional number of pigs per pen in the DB and TN genetic hybrids were 12 ± 2 stress in pigs (Donaldson et al., 2002; Weary et al., 2008) and constitutes and 13 ± 2, respectively. No pigs were tail-docked or tooth-clipped. a suitable context for the examination of the validity of play behaviour On the day of weaning, experimental pigs were (mean ± SD) 26 ± 1 as an animal welfare indicator. Accordingly, our study explored the ef­ days old and weighing 7.4 ± 1.9 kg (DB: 7.0 ± 2.0 kg; TN: 7.8 fects of weight gain, feeding behaviour (i.e., visits to feeder before and ± 1.8 kg). At weaning, the sow was removed out of sensory range of the after weaning and latency to visit the feeder after weaning) and drinking pigs. To accommodate all-in all-out farrowing for hygiene purposes, behaviour (i.e., visits to drinker before and after weaning and latency to weaning took place on one day per week, as close as possible to the day visit the drinker after weaning) on two types of play behaviour (loco­ when litters were 28 days old. In each batch, 1/3 of the DB and TN litters motor-rotational and social play in weaner pigs. Before and after stayed in the farrowing pen as an intact litter after weaning (13 ± 2 weaning, we hypothesised that pigs with higher weight gain and dis­ pigs/pen) and the other 2/3 were moved and mixed with another litter playing higher frequency of visits to the feeder and drinker, signalling in one of four identical 5.40 × 2.45 m two-climate weaner pens (26 ± 2 better nourishment, would show a higher engagement with play pigs/pen; approximately 0.51 m2/pig) with approximately 67% slatted behaviour. and 33% solid floor in the same weaner unit. The solid floor area was roofed by a manually adjustable fibre panel positioned approximately 2. Materials and methods 85 cm above the floor level (Jyden A/S, Denmark) providing two zones in the pen (a cooler slatted and drained floor area and a warmer and This study took place from February to November 2020 at the darker area under the roof). The drinker was the same type as the one experimental facilities at the Department of Animal and Veterinary used in the farrowing pens. Two identical weaner feeders (TR4, Sciences, Aarhus University, Viborg, Denmark. The experimental pro­ Rotecna®, Spain; Fig. 1B) were placed next to each other in the partially cedures were carried out in accordance with the Ministry of Food, slatted floor area, providing eight 14 × 18 cm feeding places per weaner Agriculture and Fisheries, The Danish Veterinary and Food Adminis­ pen. In both locations (farrowing pen and two-climate weaner pen), the tration under act 474 of 15. May 2014 and executive order 2028 of 14. pigs were ad libitum fed a pelleted standardised weaner diet (Prime Midi December 2020, and under consideration of the ARRIVE Guidelines (Du Piller, DLG, Denmark, metabolizable energy: 14.8 MJ/kg, 19.3% crude Sert et al., 2020). The present study is part of a large study examining the protein), with no medical zinc oxide added to it, delivered once daily in clinical, physiological, and behavioural effects of weaning intact litters the morning. The room temperature was increased to 24 ◦ C and grad­ in farrowing pens for loose-housed sows and sow hybrid featured by ually decreased to 21 ◦ C 56 days post-weaning. At the time of weaning reduced litter size and pigs with increased individual birth and weaning and every morning post-weaning, all farrowing pens received approxi­ weight described in detail by Franchi et al., (2022b) and Winters et al. mately 130 g of chopped wheat straw on the solid floor and approxi­ (2022). mately 400 g of sawdust in the creep area. The two-climate weaner pens received double the amount of straw and sawdust (to maintain same 2.1. Animals, housing, and management provision per animal) placed on the partly slatted floor and on the solid floor, respectively. Six DanBred Landrace × Yorkshire (DB) gilts and six Topigs Norsvin 2 G.A. Franchi et al. Applied Animal Behaviour Science 259 (2023) 105836 Fig. 1. Illustration of the feeders and drinker installed in the farrowing pen (subfigure A) and weaner pen (subfigure B). Both farrowing and two-climate weaner pens were equipped with similar drinkers. 2.2. Clinical recording 2.3.2. Feeding and drinking behaviour Individual visits to feeder and drinker were registered on days − 1 All pigs were individually weighed within 24 h of birth and on days and 1 using instantaneous sampling at 2-min intervals (Bateson and − 7, 0, 1, and 2, between 08:00 h and 12:00 h, relative to the day of Martin, 2021) for a 13-h period from 07:00 h to 20:00 h (450 scans/­ weaning. Pig weight gain in the pre-weaning period (i.e., from birth to day). This sampling interval was chosen based on time availability and weaning, from day − 7 to weaning) and post-weaning period (i.e., from confirmed by a pilot study with 13 non-experimental pigs during 13 h on weaning to day 1, from weaning to day 2, and from day 1 to day 2) was the day before weaning demonstrating the individual proportion of calculated as end period weight minus start period weight and divided visits to feeder and drinker labelled instantaneously at 2-min interval by the number of days. Weight gain is presented in grams per day (g/d). highly agreed with the respective outcomes labelled continuously (nonparametric concordance test index = 0.96 (visits to feeder) and 0.89 2.3. Behaviour recording (visits to drinker); Winters et al., unpublished data). Visits to each resource were defined as a pig having its head including both ears within A 2D video camera (model DS-2CD2145FWD-I, Hikvision, China) the area of the feeder or drinker with its nose pointing downwards was placed above each pen with a top-view angle, overlooking the entire (Winters et al., 2022). The visits to feeder and drinker were observed by pen area, except under the creep area in the farrowing pen and under the one trained experimenter and the order of observed pens was rando­ cover in the climate pen, from a height of approximately 2.3 m in the mised across treatments, batches, and days to avoid systematic drift in farrowing pen and 2.8 m in the weaner pen. Video data were processed the observation over time. Additionally, the individual latency to visit with the software Blue Iris v.5 (Blue Iris Security, USA). Each pig was the feeder and drinker (i.e., the pig having its head and both ears above identified on the video recordings by use of markings on the back made the edge of the trough with its nose pointing downward for minimum 5 s with a livestock marker on days − 2, 0, 1 and 2 at between 08:00 h and (feeder) or 2 s (drinker) (Winters et al., 2022)) within the first 24 h 12:00 h. post-weaning was recorded continuously by two experimenters and the order of observed pens was randomised between treatments and 2.3.1. Play behaviour batches. The time point was recorded to the nearest minute and a Play behaviour was observed at individual level from video re­ maximum latency of 1440 min (24 h) was used, after which latencies cordings on days − 1, 1 and 2 relative to weaning from 14:00 h to were right-censored. No intra- or inter-experimenter agreement calcu­ 22:00 h (i.e., period with least disturbance across pens and artificial lations were performed for visit measurements or latency light on) following behavioural sampling and continuous recording measurements. (Bateson and Martin, 2021) and using Noldus Observer XT 15 (Noldus Information Technology, The Netherlands) software package. The du­ 2.4. Statistical analyses rations (s/pig/8-h period) of two types of play behaviour were recorded separately: locomotor-rotational play (LOC) and social play (SOC), ac­ Statistical analyses were performed in R v.2.1–461 (R Core Team, cording to the ethogram described in Table 1. The durations of LOC and 2021) at individual level accounting for the nested structure in data with SOC of the individual pig comprised the sum of durations per 8-h pigs nested in pens. P-values < 0.05 were considered significant, and observation period of all behavioural elements constituting each type values of 0.1 > P ≥ 0.05 were considered tendencies. The current of play. Whenever pigs were disturbed (i.e., a human was within the manuscript focuses on the effects of measures of nutritional status on framework of the pen, or the pigs were not in the pen, or the pigs were play. The effects of genetic hybrids and weaning methods on LOC and being moved in or out of the pen), this was noted, and no behavioural SOC are reported in Franchi et al., (2022b), while effects on the mea­ registrations were made. The duration of disturbance across pens ranged sures of nutritional status are reported in Winters et al. (2022). The ef­ from 0 to 150 s on day − 1, 0–47 s on day 1 and 0–51 s on day 2. Three fects of weight gain, proportion of visits to feeder (i.e., number of scans experimenters were assigned to register play behaviour from video re­ visiting the feeder divided by the maximum number of daily scans) and cordings of the pigs, balanced for treatments (genetic hybrid × weaning proportion of visits to drinker (i.e., number of scans visiting the drinker method), and the inter-experimenter agreement on the total durations of divided by the maximum number of daily scans) (independent variables) LOC, SOC, and combination of LOC and SOC of the individual pig were on LOC and SOC (dependent variables) were analysed with assessed with Concordance Correlation Coefficient (CCC; Lawrence and mixed-effects modelling (library glmmTMB v.1.1.2; Brooks et al., 2017) Lin, 1989). Based on the criteria proposed by Hinkle et al. (2003), we on each day (− 1, 1 and 2) separately. The description of the models for interpreted CCC > 0.8 as high agreement level. The achieved CCC for each day relative to weaning is presented in Table 2. Each independent LOC was 0.82; for SOC was 0.91; and for the combination of LOC and variable was transformed into a categorical variable on each day using SOC was 0.82. the 33% and 67% quantiles (Table 2). On day − 1 (11% zero observa­ tions), LOC was square-root-transformed to fulfil the assumptions of 3 G.A. Franchi et al. Applied Animal Behaviour Science 259 (2023) 105836 Table 1 Table 2 Ethogram of behavioural elements constituting locomotor-rotational play (LOC) Description of independent variables’ (weight gain, proportion of visits to and social play (SOC) in domesticated pigs. feeder, proportion of visits to drinker) levels and random effects included in the Play Behavioural Definition models to examine the relationships between the independent variables and type element locomotor-rotational play and social play on each day relative to weaning (day 0). Each independent variable was transformed into categorical variable on each LOCa Gambolling Gallop-like energetic running (the 2 forelimbs move day using the 33% and 67% quantiles. Each row represents a different model. approximately in phase, followed by the 2 hind limbs) and hopping in forward motions within the Day Independent variable Covariate Random effects pen environment. Often associated with vigorous ear -1 Weight gain from birth to weaning day Birth Pen before flapping, moving across a large area of the pen, and (low: ≤209 g/d, n = 99; medium: weight weaning nested occasionally bouncing into other pigs. As part of a 209 0.1). CI [0.03–0.3]; χ22 = 17.07; P < 0.001; Fig. 4C). Pigs performing LOC on day 1 did not differ from pigs not performing LOC on day 1 regarding 3.1. Weight gain and play latency to visit the feeder within the first 24 h from weaning (χ21 = 0.24; P = 0.618). Performance of LOC on day 2 was associated with a ten­ Pigs in the low weight gain category from birth to weaning spent the dency for a sooner visit to the feeder within 24 h following weaning shortest time performing LOC on day − 1 (χ22 = 15.81; P < 0.001; (Hazard Rate Ratio (HRR) = 1.3, 95% CI [1.0–1.7]; χ21 = 3.62; Fig. 2 A). Pigs in the low weight gain category from day − 7 to weaning P = 0.057). spent the shortest time performing LOC on day − 1 (χ22 = 10.58; No effect of proportion of visits to the feeder on SOC was detected on P = 0.005; Fig. 2B). No effect of weight gain from weaning to day 1 on any of the investigated days (SOCday-1 vs. Visitsday-1: χ22 = 0.60, LOC on day 1 was detected (χ22 = 2.35; P = 0.308; Fig. 2C). Pigs in the P = 0.739; SOCday1 vs. Visitsday1: χ22 = 2.03, P = 0.363; SOCday2 vs. high (Odds Ratio (OR) = 3.9, 95% Confidence Interval (CI) [1.5–9.8]) Visitsday1: χ22 = 4.10, P = 0.128). Pigs not performing SOC did not differ and medium (OR = 2.8, 95% CI [1.2–6.4]) weight gain categories from from pigs performing SOC on day 1 (χ21 = 1.02; P = 0.313) and day 2 (χ21 weaning to day 2 showed higher odds of performing LOC on day 2 than = 1.14; P = 0.287) in latency to visit the feeder within 24 h from pigs in the low weight gain category (χ22 = 11.79; P = 0.003; Fig. 2D). weaning. Pigs in the high weight gain category from days 1–2 showed higher odds of performing LOC on day 2 than pigs in the low weight gain category 3.3. Drinking behaviour and play (OR = 2.9, 95% CI [1.2–7.1]; χ22 = 6.46; P = 0.039; Fig. 2E). Pigs in the low weight gain category from birth to weaning showed Pigs with low proportion of visits to the drinker on day − 1 spent the the lowest odds of performing SOC on day − 1 (ORL/M = 0.2, 95% CI shortest time performing LOC on day − 1 (χ22 = 13.26; P = 0.001; [0.1–0.6]; ORL/H = 0.3, 95% CI [0.1–0.7]; χ22 = 12.79; P = 0.002; Fig. 5A). No effect of proportion of visits to the drinker on day 1 on LOC Fig. 3A). Pigs in the low weight gain category from day − 7 to weaning on day 1 was detected (χ22 = 1.42; P = 0.492; Fig. 5B). Pigs with low showed the lowest odds of performing SOC on day − 1 than pigs in the proportion of visits to the drinker on day 1 showed higher odds of low weight gain category (ORL/M = 0.4, 95% CI [0.2–0.8]; ORL/H = 0.4, performing LOC on day 2 than pigs with medium proportion of visits to 95% CI [0.1–1.0]; χ22 = 6.47; P = 0.039; Fig. 3B). Pigs in the high weight the drinker on day 1 (OR = 3.2, 95% CI [1.4–7.1]; χ22 = 8.83; P = 0.012; gain category from weaning to day 1 showed higher odds of performing Fig. 5C). Pigs not performing LOC did not differ from pigs performing SOC on day 1 than pigs in the medium weight gain category (OR = 16.9, LOC on day 1 in latency to visit the drinker within 24 h from weaning (χ21 95% CI [1.5–192]; χ22 = 8.64; P = 0.013; Fig. 3C). No effect of weight = 0.75; P = 0.387). Performance of LOC on day 2 was associated with a gain from weaning to day 2 on SOC on day 2 was detected (χ22 = 3.63; sooner visit to the feeder within 24 h following weaning (Hazard Rate P = 0.163; Fig. 3D). Pigs in the high weight gain category from days Ratio (HRR) = 1.4, 95% CI [1.1–1.7]; χ21 = 4.06; P = 0.044). 1–2 tended to show the highest odds of performing SOC on day 2 (ORH/L No effect of proportion of visits to the drinker on SOC was detected = 3.1, 95% CI [0.5–19.5]; ORH/M = 5.4, 95% CI [1.0–28.7]; χ22 = 4.99; on any of the investigated days (SOCday-1 vs. Visitsday-1: χ22 = 4.16, P = 0.082; Fig. 3E). P = 0.125; SOCday1 vs. Visitsday1: χ22 = 1.48, P = 0.477; SOCday2 vs. Visitsday1: χ22 = 0.31, P = 0.859). Pigs not performing SOC did not differ from pigs performing SOC on day 1 (χ21 = 0.01; P = 0.928) and day 2 (χ21 3.2. Eating behaviour and play = 0.89; P = 0.344) in latency to visit the drinker within 24 h from weaning. Pigs with medium proportion of visits to the feeder on day − 1 ten­ ded to spend the longest time performing LOC on day − 1 (χ22 = 5.16; P = 0.075; Fig. 4A). Pigs with low proportion of visits to the feeder on day 1 showed the lowest odds of performing LOC on day 1 (ORL/M = 0.3, 95% CI [0.1–0.7]; ORL/H = 0.2, 95% CI [0.1–0.7]; χ22 = 8.62; P = 0.013; Fig. 2. Relationship between weight gain (A: from birth to weaning; B: from day − 7 to weaning; C: from weaning to day 1; D: from weaning to day 2; E: from days 1–2) and locomotor-rotational play (LOC) registered on days − 1 (A and B), 1 (C) and 2 (D and E) relative to weaning between 14:00 h and 22:00 h. The weight gain categories were low (L; orange bars), medium (M; blue bars), and high (H; grey bars). In sub-Figs. A and B, bars indicate least squares means of durations of LOC (s/pig/8-h) and error bars indicate standard error, square-root-transformed prior to analysis. The back-transformed least squares means are shown inside white boxes within each bar. In sub-figures C, D and E, bars indicate probability of displaying LOC and error bars indicate standard error, both back-transformed from the model using the inverse link function. Different superscript letters indicate statistical difference at P < 0.05 within days. 5 G.A. Franchi et al. Applied Animal Behaviour Science 259 (2023) 105836 Fig. 3. Relationship between weight gain (A: from birth to weaning; B: from day − 7 to weaning; C: from weaning to day 1; D: from weaning to day 2; E: from days 1–2) and probability of displaying social play (SOC) on each day relative to weaning (days − 1 (A and B), 1 (C) and 2 (D and E)) between 14:00 h and 22:00 h. The weight gain cate­ gories were low (L; orange bars), medium (M; blue bars), and high (H; grey bars). The bars represent probability, and the error bars indicate standard error. Estimates and stan­ dard error bars were back-transformed from the model using the inverse link function. Different superscript letters indicate statistical difference at P < 0.05 within days. Fig. 4. Relationship between proportion of visits to the feeder (A: day − 1; B and C: both day 1) and locomotor- rotational play (LOC) registered on days − 1 (A), 1 (B) and 2 (C) relative to weaning between 14:00 h and 22:00 h. The proportion of visits to feeder categories were low (L; red bars), medium (M; pink bars), and high (H; dark blue bars). In sub-figure A, bars indicate least squares means of durations of LOC (s/pig/8-h) and error bars indicate stan­ dard error, square-root-transformed prior to analysis. The back-transformed least squares means are shown inside white boxes within each bar. In sub-Figs. B and C, bars indicate probability of displaying LOC and error bars indicate standard error, both back-transformed from the model using the inverse link function. Different superscript letters indicate statistical difference at P < 0.05 within days. 4. Discussion showed a dramatic decline in time spent playing (Franchi et al., 2022b), illustrating the negative effects of early, abrupt weaning (i.e., separation 4.1. Play behaviour and weight gain from the sow, environmental change, and social stress). This may explain the absence of a clear relationship between LOC and weight gain We found that weight gain, which is indicative of well-nourished during this day. Given that animals are motivated to play when their individuals, positively related to LOC before and after weaning, except basic needs, such as food, are met (Boissy et al., 2007; Held and Špinka, in the first 24 h post weaning. In the current study, the weight gain from 2011), the lower motivation to perform LOC by the pigs losing weight birth to weaning reflected pigs’ development (long-term) and the weight following weaning was likely due to undernourishment and a motiva­ gain from day − 7 to weaning reflected pigs’ nutritional status pre- tional shift toward feeding and conserving body energy reserves (Bur­ weaning (short-term). Our results agree with previous studies exam­ ghardt, 1984; Jensen et al., 2015). Even though play behaviour typically ining play behaviour in weaner pigs. For instance, Brown et al. (2015) comprises a small proportion of an animal’s activity time budget (Martin found a positive relationship between weight gain and LOC and Caro, 1985), the performance of this behaviour, known to involve a pre-weaning. Moreover, Šilerová et al. (2010) found that LOC on day 5 high degree of energy expenditure (Müller-Schwarze et al., 1982), may post-weaning and weight gain during the 5 days following weaning were be sacrificed when body energy reserves or food supply are low (Som­ positively related. Within the first 24 h following weaning, all three mer and Mendoza-Granados, 1995; Nunes et al., 2004). Additionally, we weight gain categories displayed weight loss (Table 2) and all pigs cannot exclude that the pigs’ motivation may also have shifted toward 6 G.A. Franchi et al. Applied Animal Behaviour Science 259 (2023) 105836 Fig. 5. Relationship between proportion of visits to the drinker (A: day − 1; B and C: both day 1) and locomotor- rotational play (LOC) registered on days − 1 (A), 1 (B) and 2 (C) relative to weaning between 14:00 h and 22:00 h. The proportion of visits to drinker categories were low (L; purple bars), medium (M; green bars), and high (H; yellow bars). In sub-figure A, bars indicate least squares means of durations of LOC (s/pig/8-h) and error bars indicate stan­ dard error, square-root-transformed prior to analysis. The back-transformed least squares means are shown inside white boxes within each bar. In sub-Figs. B and C, bars indicate probability of displaying LOC and error bars indicate standard error, both back-transformed from the model using the inverse link function. Different superscript letters indicate statistical difference at P < 0.05 within days. exploring the novel environment and defending social status during the (Table 2). The solid feed available pre-weaning was sow lactation diet, establishment of a novel hierarchy (D’Eath, 2005; Weary et al., 2008). which was not intended to be a primary source of feed for the piglets but The relationship between SOC and weight gain was the clearest in rather the possibility of visiting the sow feeder should be seen as a social the pre-weaning period: pigs in the high and medium weight gain cat­ learning facilitator, so that pigs could learn to ingest solid feed from egories had higher odds of performing SOC than pigs in the low weight their mother (Oostindjer et al., 2011) to facilitate solid feed intake post gain category. Similar findings were reported by Brown et al. (2015) in weaning. Before weaning, pigs displaying a low frequency of visits to the pigs during the pre-weaning period and by Nunes et al., (1999, 2004) in drinker spent the shortest time performing LOC. Water intake is an ground squirrels (Spermophilus beldingi). Like LOC, SOC involves a intrinsic aspect of the nutrition, health, and welfare of pigs (Andersen considerable degree of energy expenditure in addition to requiring a et al., 2014), and variations in water intake can indicate ill health socially stable environment, which was likely achieved throughout the (Pijpers et al., 1991) or thermal stress (Pearce et al., 2014). As play weeks pre-weaning, but disrupted in the first days following weaning, behaviour involves energy and nutrient expenditure (Müller-Schwarze and especially, in the pigs that were moved and socially mixed in a novel et al., 1982), a higher drinking frequency, and potentially higher water environment (Newberry et al., 1988; Brown et al., 2015). Therefore, the intake, could have been a way to restore water and nutrients expended positive relationship between SOC and weight gain pre-weaning was during play (Bigelow and Houpt, 1988). expected. Conversely, no clear relationship between SOC and weight We found that pigs showing medium and high frequency of visits to gain post-weaning was observed. Likewise, Brown et al. (2018) reported the feeder on day 1 showed higher odds of displaying LOC on the two no relationship between SOC and weight gain in pigs after weaning. As days after weaning than pigs showing low frequency of visits to the demonstrated by Franchi et al., (2022b), accounting for the current feeder on day 1. Additionally, the performance of LOC on day 2 was study’s population, the performance of both LOC and SOC decreased associated with a shorter latency to visit to the feeder and the drinker on considerably during the first 48 h following weaning compared to the the previous day. These results suggest that those pigs that were able to pre-weaning levels. Additionally, SOC was defined herein as two or compensate the loss of milk at weaning by visiting these resources more pigs mutually engaging in, at least, one of the behaviours consti­ sooner and increasing their intake of solid feed and water were better tuting SOC, while in other studies this mutuality was not a prerequisite nourished and thus more motivated to play, supporting the findings on (e.g., Hötzel et al., 2010; Camerlink et al., 2021). Hence, the combina­ the relation between weight gain and LOC. Winters et al. (2022) also tion of the negative effect of abrupt and early weaning (e.g., regrouping, demonstrated that pigs showing shorter latency to visit these resources novel environment; Donaldson et al., 2002) and a stricter SOC defini­ were the ones with higher pre-weaning weight gain and higher visits to tion, resulting in a low level of SOC, may explain why the present study the feeder on day − 1. found no clear relationship between SOC and weight gain post-weaning. In our study, all pigs were subjected to some degree of novelty after As it is unclear whether LOC and SOC share similar motivation and weaning. Pigs kept in the farrowing pen dealt with the weaning from the whether they are similarly affected by contextual changes, we advise sow and introduction of a novel feed. Pigs moved to a weaner pen future studies to investigate the motivations for each type of play and additionally experienced a new environment (including new feeder and how this is influenced by nutritional, as well as social and health, status. drinker) and new peers. As discussed above, the motivations for each type of play behaviour likely vary beyond the individual nutritional status. This may further explain the lack of clear relationships between 4.2. Play behaviour and feeding and drinking behaviour feeding and drinking behaviour and SOC. Irrespective, in the present study, the outcomes of the analyses of the effects of the nutritional To our knowledge, this is the first study examining the relationship measures on the two types of play behaviour must be interpreted between play behaviour and measures of feeding behaviour in pigs. In cautiously, due to the low daily proportions of visits to each resource the current study, no clear relationship between play behaviour and registered in each category (Table 2) as well as the low levels of each visits to the feeder was observed before weaning. We speculate that this type of play behaviour, particularly after weaning. We encourage more finding is due to pigs being able to fulfil their nutritional requirements studies to assess the influence of other aspects than nutrition, such as by suckling the sow, resulting in a low frequency of visits to the feeder 7 G.A. Franchi et al. Applied Animal Behaviour Science 259 (2023) 105836 presence of illness, on the variation in LOC and SOC in pigs around Kjærulff Christensen and Dines Thøger Bolt (both AU) for assistance with weaning. pig handling and behavioural observations. This manuscript is devel­ oped as part of a project funded by the European Union’s Horizon 2020 4.3. Value of play as a positive animal welfare measure research and innovation programme under grant agreement No. 862919. The data collection as well as the writing of this manuscript was In our study, the performance of play behaviour was positively further funded by the Ministry of Food, Agriculture and Fisheries of related to the display of clinical and behavioural signs of being well- Denmark under the program “Veterinary agreement III” (Veter­ nourished. This association with the lack of a welfare threat (e.g., un­ inærforlig III, AU6), and by the Marie Skłodowska-Curie Action grant dernourishment or hunger) and a putative positive experience (e.g., agreement No. 842555 (project AutoPlayPig). satiation) means that increased play behaviour may be an indicator of increased welfare and as such as a positive welfare indicator (Held and References Špinka, 2011). However, we acknowledge that this covers a range of the welfare spectrum, and in our study, play behaviour cannot be inter­ Ahloy-Dallaire, J., Espinosa, J., Mason, G., 2018. Play and optimal welfare: does play indicate the presence of positive affective states. Behav. Process. 156, 3–15. https:// preted as an indicator of absolute positive affective state because we are doi.org/10.1016/j.beproc.2017.11.011. not able to distinguish between the pigs who were merely coping with Andersen, H.L., Dybkjær, L., Herskin, M.S., 2014. Growing pigs’ drinking behaviour: weaning from the pigs who were truly thriving, if we can speak of number of visits, duration, water intake and diurnal variation. Animal 8, 1881–1888. https://doi.org/10.1017/S175173111400192X. thriving during an abrupt, early weaning (Ahloy-Dallaire et al., 2018; Bateson, M., Martin, P., 2021. Measuring Behaviour: An Introductory Guide, fourth ed. Maple and Bloomsmith, 2018). To enable the experience of true positive Cambridge University Press, Cambridge, UK. https://doi.org/10.1017/ affective states in animals, appropriate nutrition, as well as social and 9781108776462. Bateson, P., Mendl, M., Feaver, J., 1990. Play in the domestic cat is enhanced by physical environment, are required (Franchi et al., 2022b; Maple and rationing of the mother during lactation. Anim. Behav. 40, 514–525. https://doi. Bloomsmith, 2018). This study shows that animals experiencing a more org/10.1016/S0003-3472(05)80532-9. stable condition (i.e., by being better nourished) are motivated to Bigelow, J.A., Houpt, T.R., 1988. Feeding and drinking patterns in young pigs. Physiol. Behav. 43, 99–109. https://doi.org/10.1016/0031-9384(88)90104-7. perform more play behaviour. For play behaviour to indicate the pres­ Boissy, A., Manteuffel, G., Jensen, M.B., Moe, R.O., Spruijt, B., Keeling, L.J., Winckler, C., ence of positive affective states, it should increase when positive con­ Forkman, B., Dimitrov, I., Langbein, J., Bakken, M., 2007. Assessment of positive ditions become even more positive (Ahloy-Dallaire et al., 2018), and this emotions in animals to improve their welfare. Physiol. Behav. 92, 375–397. https:// doi.org/10.1016/j.physbeh.2007.02.003. should also be demonstrated in other (more) positive contexts than Brooks, M.E., Kristensen, K., van Benthem, K.J., Magnusson, A., Berg, C.W., Nielsen, A., weaning. Hence, we advise continuation of the validation of play against Skaug, H.J., Maechler, M., Bolker, B.M., 2017. glmmTMB balances speed and physiological (e.g., stress hormones, peripheral body temperature), flexibility among packages for zero-inflated generalized linear mixed modeling. R. J. clinical (e.g., presence of disease or injuries), behavioural (e.g., body 9, 378–400. https://doi.org/10.3929/ethz-b-000240890. Brown, S.M., Klaffenböck, M., Nevison, I.M., Lawrence, A.B., 2015. Evidence for litter postures), and cognitive measures (e.g., cognitive bias, reward loss differences in play behaviour in pre-weaned pigs. Appl. Anim. Behav. Sci. 172, sensitivity) (as reviewed by Kremer et al., 2020) to understand and 17–25. https://doi.org/10.1016/j.applanim.2015.09.007. confirm the value of play behaviour as a positive animal welfare Brown, S.M., Peters, R., Nevison, I.M., Lawrence, A.B., 2018. Playful pigs: evidence of consistency and change in play depending on litter and developmental stage. Appl. indicator. Anim. Behav. Sci. 198, 36–43. https://doi.org/10.1016/j.applanim.2017.09.018. Burgdorf, J., Panksepp, J., 2007. The neurobiology of positive emotions. Neurosci. 5. Conclusions Biobehav. Rev. 30, 173–187. https://doi.org/10.1016/j.neubiorev.2005.06.001. Burghardt, G.M., 1984. On the origins of play. In: Smith, P.K. (Ed.), Play in Animals and Humans. Blackwell, Oxford, pp. 5–41. Our results support the hypothesis that animals experiencing a more Burghardt, G.M., 2005. The Genesis of Animal Play: Testing the Limits. MIT Press. stable condition (e.g., better nourishment) are more motivated to Camerlink, I., Proßegger, C., Kubala, D., Galunder, K., Rault, J.L., 2021. Keeping littermates together instead of social mixing benefits pig social behaviour and perform play behaviour. The current study represents an advancement growth post-weaning. Appl. Anim. Behav. Sci. 235, 105230 https://doi.org/ in the understanding and validation of play behaviour as an animal 10.1016/j.applanim.2021.105230. welfare indicator. D’Eath, R.B., 2005. Socialising piglets before weaning improves social hierarchy formation when pigs are mixed postweaning. Appl. Anim. Behav. Sci. 93, 199–211. https://doi.org/10.1016/j.applanim.2004.11.019. CRediT authorship contribution statement Donaldson, T.M., Newberry, R.C., Špinka, M., Cloutier, S., 2002. Effects of early play experience on play behaviour of piglets after weaning. Appl. Anim. Behav. Sci. 79, Guilherme A. Franchi: Conceptualization, Investigation, Method­ 221–231. https://doi.org/10.1016/S0168-1591(02)00138-7. Du Sert, N.P., Ahluwalia, A., Alam, S., Avey, M.T., Baker, M., Browne, W.J., Clark, A., ology, Data curation, Visualization, Formal analysis, Software. Mona L. Cuthill, I.C., Dirnagl, U., Emerson, M., Garner, P., 2020. Reporting animal research: V. Larsen: Conceptualization, Investigation, Methodology, Funding explanation and elaboration for the ARRIVE guidelines 2.0. PLoS Biol. 18, e3000411 acquisition, Resources, Project administration, Supervision. Ida H. https://doi.org/10.1371/journal.pbio.3000411. Duve, L.R., Weary, D.M., Halekoh, U., Jensen, M.B., 2012. The effects of social contact Kristoffersen: Data curation, Visualization. Jeanet F.M. Winters: and milk allowance on responses to handling, play, and social behavior in young Conceptualization, Investigation, Methodology. Lene Juul Pedersen: dairy calves. J. Dairy Sci. 95, 6571–6581. https://doi.org/10.3168/jds.2011-5170. Conceptualization, Investigation, Methodology, Funding acquisition, Franchi, G.A., Larsen, M.L.V., Kristoffersen, I.H., Winters, J.F.M., Jensen, M.B., Pedersen, L.J., 2022a. Supplemental material: play behaviour positively relates with Resources, Project administration, Supervision. Margit Bak Jensen: growth, feeding behaviour and drinking behaviour in weaner pigs (Sus scrofa). Conceptualization, Investigation, Methodology, Funding acquisition, Mendeley Data v1. 10.17632/n6hsg8kcjf.1. Resources. All authors contributed to the article and approved the Franchi, G.A., Larsen, M.L.V., Winters, J.F.M., Jensen, M.B., Pedersen, L.J., 2022b. Investigating the effects of two weaning methods and two genetic hybrids on play submitted version. behaviour in weaner pigs (Sus scrofa). Front. Anim. Sci. 3, 909038 https://doi.org/ 10.3389/fanim.2022.909038. Declaration of Competing Interest Hall, S.L., Bradshaw, J.W., 1998. The influence of hunger on object play by adult domestic cats. Appl. Anim. Behav. Sci. 58, 143–150. https://doi.org/10.1016/ S0168-1591(97)00136-6. The authors declare that they have no known competing financial Held, S.D., Špinka, M., 2011. Animal play and animal welfare. Anim. Behav. 81, interests or personal relationships that could have appeared to influence 891–899. https://doi.org/10.1016/j.anbehav.2011.01.007. the work reported in this paper. Hinkle, D.E., Wiersma, W., Jurs, S.G., 2003. Applied Statistics for the Behavioural Sciences, hird ed. Houghton Mifflin,, Boston, MA, USA. Hothorn, T., Bretz, F., Westfall, P., 2008. Simultaneous inference in general parametric Acknowledgments models. Biom. J. 50, 346–363. https://doi.org/10.1002/bimj.200810425. Hötzel, M.J., Machado Filho, L.C.P., Irgang, R., Alexandre Filho, L., 2010. Short-term behavioural effects of weaning age in outdoor-reared piglets. Animal 4, 102–107. The authors acknowledge the help from the staff at the Pig Research https://doi.org/10.1017/S1751731109990875. Unit, Aarhus University (AU-Viborg), Denmark. We also thank Carsten 8 G.A. Franchi et al. Applied Animal Behaviour Science 259 (2023) 105836 Jensen, M.B., Duve, L.R., Weary, D.M., 2015. Pair housing and enhanced milk allowance Pearce, S.C., Sanz-Fernandez, M.V., Hollis, J.H., Baumgard, L.H., Gabler, N.K., 2014. increase play behavior and improve performance in dairy calves. J. Dairy Sci. 98, Short-term exposure to heat stress attenuates appetite and intestinal integrity in 2568–2575. https://doi.org/10.3168/jds.2014-8272. growing pigs. J. Anim. Sci. 92, 5444–5454. https://doi.org/10.2527/jas.2014-8407. Jensen, P., Stangel, G., 1992. Behaviour of piglets during weaning in a seminatural Pijpers, A., Schoevers, E.J., Van Gogh, H., van Leengoed, L.A.M.G., Visser, I.J.R., van enclosure. Appl. Anim. Behav. Sci. 33, 227–238. https://doi.org/10.1016/S0168- Miert, A.S.J.P.A.M., Verheijden, J.H.M., 1991. The influence of disease on feed and 1591(05)80010-3. water consumption and on pharmacokinetics of orally administered oxytetracycline Lawrence, I., Lin, K., 1989. A concordance correlation coefficient to evaluate in pigs. J. Anim. Sci. 69, 2947–2954. https://doi.org/10.2527/1991.6972947x. reproducibility. Biometrics 255–268. https://doi.org/10.2307/2532051. R Core Team. (2021). R: A language and environment for statistical computing. Vienna: Lenth, R.V. (2021). emmeans: estimated marginal means, aka least-squares means. R R Foundation for Statistical Computing. Available online at: 〈https://www.R-project. package version 1.6.3. 〈https://CRAN.R-project.org/package=emmeans〉 (accessed org/〉 (accessed 01 June 2022). 01 June 2022). Rushen, J., Wright, R., Johnsen, J.F., Mejdell, C.M., de Passillé, A.M., 2016. Reduced Loranca, A., Torrero, C., Salas, M., 1999. Development of play behavior in neonatally locomotor play behaviour of dairy calves following separation from the mother undernourished rats. Physiol. Behav. 66, 3–10. https://doi.org/10.1016/S0031- reflects their response to reduced energy intake. Appl. Anim. Behav. Sci. 177, 6–11. 9384(98)00235-2. https://doi.org/10.1016/j.applanim.2016.01.023. Loy, J., 1970. Behavioral responses of free-ranging rhesus monkeys to food shortage. Am. Sharpe, L.L., Clutton-Brock, T.H., Brotherton, P.N., Cameron, E.Z., Cherry, M.I., 2002. J. Phys. Anthropol. 33, 263–271. https://doi.org/10.1002/ajpa.1330330212. Experimental provisioning increases play in free-ranging meerkats. Anim. Behav. 64, Maple, T.L., Bloomsmith, M.A., 2018. Introduction: the science and practice of optimal 113–121. https://doi.org/10.1006/anbe.2002.3031. animal welfare. Behav. Process. 156, 1–2. https://doi.org/10.1016/j. Šilerová, J., Špinka, M., Šárová, R., Algers, B., 2010. Playing and fighting by piglets beproc.2017.09.012. around weaning on farms, employing individual or group housing of lactating sows. Martin, P., Caro, T.M., 1985. On the functions of play and its role in behavioral Appl. Anim. Behav. Sci. 124, 83–89. https://doi.org/10.1016/j. development. In: Advances in the Study of Behavior, Vol. 15. Academic Press, applanim.2010.02.003. pp. 59–103. https://doi.org/10.1016/S0065-3454(08)60487-8. Siviy, S.M., Bekoff, M., Beyers, J.A., 1998. Neurobiological substrates of play behavior. Miguel-Pacheco, G.G., Vaughan, A., de Passillé, A.M., Rushen, J., 2015. Relationship Animal play. Evolutionary, Comparative and Ecological Perspectives. Cambridge between locomotor play of dairy calves and their weight gains and energy intakes University Press, Cambridge, UK, pp. 221–242. around weaning. Animal 9, 1038–1044. https://doi.org/10.1017/ Sommer, V., Mendoza-Granados, D., 1995. Play as indicator of habitat quality: a field S1751731115000063. study of langur monkeys (Presbytis entellus. Ethology 99, 177–192. https://doi.org/ Müller-Schwarze, D., Stagge, B., Muller-Schwarze, C., 1982. Play behavior: persistence, 10.1111/j.1439-0310.1995.tb00893.x. decrease, and energetic compensation during food shortage in deer fawns. Science Špinka, M., 2017. In: Jensen, P. (Ed.), “Behaviour of Pigs” in the Ethology of Domestic 215, 85–87 science.215.4528.85. Animals, third ed. CAB International Publishing, pp. 214–227. Newberry, R.C., Wood-Gush, D.G.M., Hall, J.W., 1988. Playful behaviour of piglets. Špinka, M., Newberry, R.C., Bekoff, M., 2001. Mammalian play: training for the Behav. Process. 17, 205–216. https://doi.org/10.1016/0376-6357(88)90004-6. unexpected. Q. Rev. Biol. 76, 141–168. https://doi.org/10.1086/393866. Nunes, S., Muecke, E.M., Anthony, J.A., Batterbee, A.S., 1999. Endocrine and energetic Therneau, T. (2020a). A package for survival analysis in R. R package version 3.2–3. 〈htt mediation of play behavior in free-living Belding’s ground squirrels. Horm. Behav. ps://CRAN.R-project.org/package=survival〉 (accessed 01 June 2022). 36, 153–165. https://doi.org/10.1006/hbeh.1999.1538. Therneau, T.M. (2020b). coxme: mixed effects Cox models. R package version 2.2–16. Nunes, S., Muecke, E.M., Lancaster, L.T., Miller, N.A., Mueller, M.A., Muelhaus, J., 〈https://CRAN.R-project.org/package=coxme〉 (accessed 01 June 2022). Castro, L., 2004. Functions and consequences of play behaviour in juvenile Belding’s Weary, D.M., Jasper, J., Hötzel, M.J., 2008. Understanding weaning distress. Appl. Anim. ground squirrels. Anim. Behav. 68, 27–37. https://doi.org/10.1016/j. Behav. Sci. 110, 24–41. https://doi.org/10.1016/j.applanim.2007.03.025. anbehav.2003.06.024. Winters, J.F.M., Foldager, L., Pedersen, L.J., 2022. Postweaning growth and feeding Oostindjer, M., Bolhuis, J.E., Mendl, M., Held, S., van den Brand, H., Kemp, B., 2011. behaviour in pigs: effect of heavier hybrid and weaning intact litters in farrowing Learning how to eat like a pig: effectiveness of mechanisms for vertical social pens for loose-housed sows. Animal, 100688. https://doi.org/10.1016/j. learning in piglets. Anim. Behav. 82, 503–511. https://doi.org/10.1016/j. animal.2022.100688. anbehav.2011.05.031. Yeates, J.W., Main, D.C., 2008. Assessment of positive welfare: a review. Vet. J. 175, 293–300. https://doi.org/10.1016/j.tvjl.2007.05.009. 9