Dissertation Krugmann
Transcript of Dissertation Krugmann
Schriftenreihe des Instituts für Tierzucht und Tierhaltung der
Christian-Albrechts-Universität zu Kiel, Heft 228, 2019
©2019 Selbstverlag des Instituts für Tierzucht und Tierhaltung
der Christian-Albrechts-Universität zu Kiel
Olshausenstraße 40, 24098 Kiel
Schriftleitung: Prof. Dr. J. Krieter
ISSN: 0720-4272
Gedruckt mit Genehmigung des Dekans der Agrar- und Ernährungswissen-
schaftlichen Fakultät der Christian-Albrechts-Universität zu Kiel
Aus dem Institut für Tierzucht und Tierhaltung
der Agrar- und Ernährungswissenschaftlichen Fakultät
der Christian-Albrechts-Universität zu Kiel
Assessment of growing pigs‘ positive affective state using
behavioural parameters and structural equation modelling
Dissertation
zur Erlangung des Doktorgrades
der Agrar- und Ernährungswissenschaftlichen Fakultät
der Christian-Albrechts-Universität zu Kiel
vorgelegt von
M.Sc. Katja Lisabeth Krugmann
aus Lübeck
Kiel, 2019
Dekan: Prof. Dr. Dr. C. Henning
1. Berichterstatter: Prof. Dr. Joachim Krieter
2. Berichterstatter: Prof. Dr. Nicole Kemper
Tag der mündlichen Prüfung: 15. Mai 2019
Die Förderung erfolgte dankenswerter Weise aus Mitteln des Bundesministeriums für Ernährung und Landwirtschaft (BMEL) aufgrund eines Beschlusses des Deutschen
Bundestages und der H. WILHELM SCHAUMANN STIFTUNG.
Meiner Familie
TABLE OF CONTENTS
GENERAL INTRODUCTION ........................................................................................ 11
CHAPTER ONE Are behavioural tests capable of measuring positive affective states in growing pigs? .......... 17
CHAPTER TWO Playing behaviour – a suitable indicator to measure positive emotions in growing pigs? ..... 39
CHAPTER THREE Can tail and ear postures be suitable to capture the affective state of growing pigs? ............ 63
CHAPTER FOUR Investigation of influence of growing pigs’ positive affective state on behavioural and
physiological parameters using structural equation modelling ............................................... 85
GENERAL DISCUSSION ..................................................................................................... 111
GENERAL SUMMARY ........................................................................................................ 125
ZUSAMMENFASSUNG ....................................................................................................... 129
ATTACHMENT .................................................................................................................... 133
11
GENERAL INTRODUCTION
In the last few decades, there has been increasing concern about livestock’s husbandry and
animal welfare issues (SandØe & Simonsen, 1992). Animal welfare represents a
multidimensional complex about which there are different opinions. One well-known
definition of animal welfare includes three general components: 1) basic health and biologic
functioning, which include animals’ freedom from disease and suffering and their
fundamental capability to perform, 2) natural living, in accordance with given opportunities to
practice natural behaviour in their housing systems and 3) the animals’ affective state,
especially regarding the share of their experienced positive and negative emotions (Fraser,
2008). As all three components are equally important, good welfare is only achieved if all
components are satisfied. Whereas extensive research in the area of the basic health and
functioning and natural living has already been carried out (Blokhuis, 2013), the animals’
affective state still constitutes a latent variable which is not itself measurable and must be
evaluated indirectly by different parameters. According to Fraser (2008), the animals’
affective state consists of different emotions such as pleasure, happiness, pain and suffering
and other feelings such as hunger and thirst that are experienced as pleasant or unpleasant.
Additionally, the primarily positive affective state presumably includes experienced, pleasant
emotions such as happiness (Ortony and Turner, 1990; Diener and Lucas, 2000) whereas
unpleasant emotions such as fear, pain or suffering probably indicate the animals’ more
negative affective state. In principle, emotions are described as intense, short-lived affective
responses to a stimuli related to specific body changes (Dantzer, 1988). More classically, they
are characterised as containing a behavioural component (a posture or an activity), an
autonomic component (visceral and endocrine responses) and a subjective component
(emotional experience or feeling) (Dantzer, 1988). Further, it is stated that emotional
properties do exist in animals since emotional reactivity has been identified as consistent
across time and situation (Boissy, 1998; Erhard and Schouten, 2001).
For the assessment of animals’ negative affective states, stress-related parameters such as e.g.
salivary cortisol concentrations are examined, while reliable measurement methods for the
animals’ positive affective state are lacking (Marcet Rius et al., 2018). Up to now, assessment
systems such as the “Welfare Quality®” protocol have been partially unreliable and strongly
influenced by subjective perceptions, especially in terms of the “Qualitative Behaviour
12
Assessment” (Czycholl et al., 2017; Temple et al., 2011; Bokkers et al., 2012; Tuyttens et al.,
2014). Thus, an objective measurement of the primarily positive affective state still remains
considerably difficult and an enormous challenge (Duncan, 2005). Hence, scientific
investigation of positive emotions has long been omitted (Boissy et al., 2007), but has
continued to gain significance in the course of current livestock sciences.
In previous studies, animals in cognitive bias tests have responded differently according to
their affective state (Harding et al., 2004; Scollo et al., 2014). Correspondingly, animals in a
more negative affective state, due to anxiety, depression or a barren environment, react with a
more pessimistic behavioural response to stimuli. Thus, it seems conceivable that behavioural
parameters such as the animals’ way of reacting to stimuli in standardised tests such as the
human-approach or novel-object test, playing behaviour or body language signals could be
suitable to draw conclusions concerning the primarily positive affective state. For instance,
playing behaviour in livestock is stated as “luxury” behaviour (Lawrence, 1987), which only
occurs when the animals’ primary needs reach a satisfactory level and optimal environmental
conditions prevail (Lawrence, 1987; Held and Špinka, 2011). Further, in human depression
research, physiological changes in hippocampal structures, adrenal glands (Czéh et al., 2006)
and salivary immunoglobulin-A contents (Bosch et al., 2004) and protein compositions
(Grigoriev et al., 2003) have proven to be indicative of more or fewer positive affective states,
which could possibly be verified in animals as well.
Aim of the present thesis
The aim of this thesis was to obtain a better understanding of livestock’s particularly positive
affective state, exemplified by investigations with fattening pigs. In the course of this,
potential indicators were to be derived which enabled feasible and reliable measurement of
the pigs’ particularly positive affective state. Various behavioural and physiological
parameters were selected from previous literature studies. All parameters were examined for
their suitability as potential indicators to assess the fattening pigs’ primarily positive affective
state in two different housing systems on three different farms.
The first Chapter investigates whether behavioural tests such as the human approach (HAT)
or novel object test (NOT), which are considered as suitable tests for assessing the level of
fear or anxiety in animals, are suitable to detect fattening pigs’ positive affective state. Each
pig was subjected three times to the HAT and NOT during fattening whereby the first, second
13
and third points of testing were at the beginning, middle and end of fattening, respectively. At
each point of testing, the pigs were tested alone in their home pen and three behavioural
variables were noticed and subsequently analysed: the approach latency (AL), duration of
contacts (DC) and number of contacts (NC).
The second Chapter compares the occurrence of the fattening pig’s playing behaviour in the
two different housing systems regarding the total duration of playing behaviour (s/h),
locomotor play (s/h) and social play (s/h) and whether these could be useful to assess the pig’s
particularly positive affective state. The playing behaviour was recorded for two days at the
beginning and two days at the end of fattening.
The third Chapter deals with body language signals of the examined fattening pigs, such as
their curled-up tails or forward-directed ears and proved their suitability to enable detection of
the pigs’ primarily positive affective state. The body language signals were analysed for four
days in total, two days at the beginning and two days at the end of fattening, respectively.
The fourth Chapter provides insights into the relationships between a large variety of
behavioural parameters such as behavioural tests, playing behaviour or body language signals
and physiological parameters such as hippocampal structures and salivary immunoglobulin-A
content respectively protein compositions which were collected during the related research
project. The partial least squares structural equation modelling method was applied to analyse
the most appropriate parameters to estimate the latent variable of the fattening pigs’ primarily
positive affective state and to evaluate the latent structures in between the behavioural and
physiological parameters.
14
REFERENCES
Blokhuis, H., Veissier, I., Jones, B., Miele, M., 2013. Improving farm animal welfare –
Science and society working together: the Welfare Quality approach. In: Blokhuis, H.,
Miele, M., Veissier, I., Jones, B. (Eds.), Improving farm animal welfare. Wageningen
Academic Publishers, Wageningen, Gelderland, Netherlands.
Boissy A., 1998. Fear and fearfulness in determining behavior. In: Grandin T, editor. Genetics
and the behavior of domestic animals. New York: Academic Press; 67–111.
Boissy, A., Manteuffel, G., Jensen, Bak, M., Randi Oppermann, M., Spruijt, B., Keeling, L.
J., 2007. Assessment of positive emotions in animals to improve their welfare. In:
Physiology & behavior 92 (3), 375–397.
Bokkers, E.A.M., de Vries, M., Antonissen, I., de Boer, I.J.M., 2012. Inter- and intra-observer
reliability of experienced and inexperienced observers for the Qualitative Behaviour
Assessment in dairy cattle. Animal Welfare 21, 307-318.
Bosch, J.A., Ring, C., Amerongen, A.V.N., 2004. Academic examinations and immunity:
academic stress or examination stress? Psychosomatic Medicine, 625–627.
Czéh, B., Simon, M., Schmelting, B., Hiemke, C., Fuchs, E., 2006. Astroglial Plasticity in the
hippocampus is affected by psychological stress and concomitant fluoxetine treatment.
Neuropsychopharmacology, 1616–1626.
Czycholl, I., Kniese, C., Schrader, L., Krieter, J., 2017. Assessment of the multi-criteria
evaluation system of the Welfare Quality® protocol for growing pigs. animal 11,
1573–1580.
Dantzer R., 1988. Les émotions. Paris: Presses Universitaires de France; 121.
Diener, E., Lucas, R.E., 2000. Subjective emotional well-being. In M. Lewis and J.M.
Haviland-Jones (Eds.), Handbook of emotions (2nd ed., pp. 325-337). New York:
Guilford Press.
Duncan, I.J.H., 2005. Science-based assessment of animal welfare: farm animals. Revue
Scientifique et Technique – Office International de Epizooties 24(2):483-492.
15
Erhard H.W., Schouten W.G.P., 2001. Individual differences and personality. In: Keeling LJ,
Gonyou HW, editors. Social behavior in farm animals. Wallingford: CAB
International; 333–52.
Fraser, D., 2008. Understanding animal welfare. Acta Veterinaria Scandinavica.
Grigoriev, I.V., Nikolaeva, L.V., Artamanov, I.D., 2003. Protein Content of Human Saliva in
Various Psycho-emotional States. Biochemistry, 405–406.
Harding, E.J., Paul, E.S., Mendl, M., 2004. Animal behaviour: cognitive bias and affective
state. Nature 427, 312.
Held, S.D.E., Špinka, M., 2011. Animal Play and animal welfare. Animal Behaviour 81, 891-
899.
Lawrence, A., 1987. Consumer demand theory and the assessment of animal welfare. Animal
Behaviour 35, S. 293-295.
Marcet Rius, M., Cozzi, A., Bienboire-Frosini, C., Teruel, E., Chabaud, C., Monneret, P.,
2018. Selection of putative indicators of positive emotions triggered by object and
social play in mini-pigs. Applied Animal Behaviour Science 202, 13–19.
Ortony, A., Turner, T.J., 1990. What’s basic about basic emotions? Psychological Review, 97,
315-331.
SandØe, P., Simonsen, H.B., 1992. Assessing animal welfare: where does science end and
philosophy begin? Animal Welfare 1, 257-267.
Scollo, A., Gottardo, F., Contiero, B., Edwards, S.A., 2014. Does stocking density modify
affective state in pigs as assessed by cognitive bias, behavioural and physiological
parameters? Applied Animal Behaviour Science 153, 26–35.
Temple, D., Dalmau, A., Ruiz de la Torre, J., Manteca, X., Velarde, A., 2011. Application of
the Welfare Quality protocol to assess growing pigs kept under intensive conditions in
Spain. Journal of Veterinary Behavior 6:138–149.
Tuyttens, F.A.M., de Graaf, S., Heerkens, J.L.T., Jacobs, L., Nalon, E., Ott, S., Stadig, L.,
Van Laer, E., Ampe, B., 2014. Observer bias in animal behaviour research: can we
believe what we score, if we score what we believe? Animal Behaviour 90, 273-280.
16
17
CHAPTER ONE
Are behavioural tests capable of measuring positive affective states in
growing pigs?
K.L. Krugmann , F.J. Warnken, J. Krieter, I. Czycholl
Institute of Animal Breeding and Husbandry, Christian-Albrechts-University Kiel,
Olshausenstr. 40, D-24098 Kiel, Germany.
Submitted to Applied Animal Behaviour Science
18
ABSTRACT
This study examined whether behavioural tests such as the human approach (HAT) or novel
object test (NOT), which are considered as suitable tests for assessing the level of fear or
anxiety in animals, are suitable to detect a positive affective state in 297 fattening pigs from
three different farms. The investigated farms consisted of a barren (farm 1, n=160) and an
enriched (farm 2, n=106; farm 3, n=31) housing system that varied in terms of availability of
space, enrichment and seasonal influences. Each pig was subjected three times to the HAT
and NOT during fattening whereby the first, second and third points of testing were at the
start, middle and end of fattening, respectively. Measuring the time (in seconds) which the
pigs needed to come into physical contact with the human or novel object resulted in three
behavioural variables: the approach latency (AL), duration of contacts (DC) and number of
contacts (NC). The pigs housed in the barren environment showed quicker AL than the
enriched-housed pigs (HAT: farm 1: 11.3 ± 1.1s vs. farm 2: 65.4 ± 1.1s respectively farm 3:
56.2 ± 1.3s (in the middle of fattening); HAT: farm 1: 7.4 ± 1.1s vs. farm 2: 57.1 ± 1.1s
respectively farm 3: 58.3 ± 1.3s (at the end of fattening); NOT: farm 1: 6.4 ± 1.1s vs. farm 2:
21.7 ± 1.1s respectively farm 3: 10.9 ± 1.2s (in the middle of fattening); NOT: farm 1: 4.5 ±
1.1s vs. farm 2: 23.0 ± 1.1s respectively farm 3: 9.0 ± 1.2s (at the end of fattening)). The same
pattern of behaviour was observed for the DC in the HAT but not in the NOT (HAT: farm 1:
48.6 ± 1.1s vs. farm 2: 4.8 ± 1.1s respectively farm 3: 4.0 ± 1.3s (in the middle of fattening);
HAT: farm 1: 83.8 ± 1.1s vs. farm 2: 6.3 ± 1.1s respectively farm 3: 7.6 ± 1.3s (at the end of
fattening)). Due to controversially discussed literature, it is difficult to conclude whether the
described definite differences of the pigs’ behaviour between the two housing systems might
indicate useful behavioural indicators to detect a positive affective state. Nevertheless, the
results of this study lay relevant foundations for further investigations concerning the
assessment of the positive affective state of fattening pigs. In addition, this study increases
knowledge regarding the validity of the HAT and NOT in general.
Keywords: animal welfare; human approach test; novel object test; pigs; positive emotions
19
INTRODUCTION
Certain behavioural reactions of animals are commonly measured by behavioural tests
performed in standardized environments and seem to capture some parts of the behavioural
tendencies of the individual animal (Forkman et al., 2007). Thus, implementations of
behavioural tests are also used in breeding decisions (Wilsson and Sundgren, 1997).
Additionally, a large number of behavioural tests focus on assessing emotions (Carreras et al.,
2017) whereas the HAT and NOT are known to be suitable to measure the level of fear or
anxiety in animals (e.g. Murphy et al., 2014; Hemsworth and Coleman, 1998). Moreover, they
are used to evaluate the effectiveness of medical treatments such as anxiolytics (Dalmau et al.,
2009). Hence, the reactions to sudden or unfamiliar events are used to assess the animals’
fearfulness (Boissy, 1995) whereby the amount of avoidance or approach behaviour is
suggested to provide a measure of the level of fear in animals (Hemsworth and Coleman,
1998). Accordingly, a shorter latency to approach novel stimuli is maintained to show less
fearful animals (Forkman et al., 2007) so that it could be imagined that these animals also
possess a more positive affective state since fear is detected as a strong negative emotion,
which is also often included in assessments of animal welfare (Forkman et al., 2007). The
affective state, which is determined as an essential part of animal welfare, constitutes the
experienced positive and negative emotions of animals (Fraser, 2009). However, it still poses
an enormous challenge to measure the affective state of animals objectively as previous
assessment systems are partially unreliable and strongly influenced by subjective perceptions
(Czycholl et al., 2017; Temple et al., 2011; Bokkers et al., 2012; Tuyttens et al., 2014). The
generally criticized intensive housing conditions of farm animals and a greater demand for
more “animal friendly” systems, the necessity to detect reliable and suitable indicators that
identify animals’ especially positive affective state (Marcet Rius et al., 2018) have continued
to gain significance in the course of the increasing interest in animal welfare.
It is conceivable that behavioural features such as the latency to approach and contact novel
stimuli, which represents the usual measure to assess fear (Carreras et al., 2017) and other
measures, such as the duration or number of contacts provide valuable information on
behavioural patterns of animals. Aside from the measuring of the level of fear, the
behavioural variables mentioned reflect the willingness to explore novel objects, and the
boldness of the animals’ personality can also be measured through responses to novel objects,
threatening stimuli or environments (Carter et al., 2012). Furthermore, van der Staay et al.
20
(2017) suggest that the affective state of animals is influenced by environmental conditions.
Moreover, primarily the NOT appears to be sensitive to marginal changes in the environment
(Bracke and Spoolder, 2008). Hence, it could be imaginable that animals which are housed in
differing environments, e.g. in terms of barren or enriched habitats, different handling
treatments or further stimuli, react in different ways to sudden or unfamiliar events as are said
to be measured in behavioural tests (Boissy, 1995). Possibly, these expected, different
behavioural reactions could be useful to obtain a better comprehension of the affective state of
farm animals.
Thus, the present study aimed at determining behavioural patterns of fattening pigs housed in
different habitats that could be measured in the HAT and the NOT to detect suitable
behavioural indicators to assess livestock’s affective state. Thereby it has been hypothesized
that particularly quicker AL and longer DC indicate less fearful and more curious pigs that
simultaneously might possess a more positive affective state.
MATERIALS AND METHODS
Animals and housing conditions
The behavioural data were collected on three farms with different housing systems in
Northern Germany between November 2016 and September 2017. 297 fattening pigs bred
from commercial crossbred dams (Large White x Landrace) and sired by Pietrain boars were
tested in total. The tails of the pigs were undocked, and the boars were castrated. The housing
systems differed primarily in terms of the space availability (m²/pig), an enriched or barren
habitat and the climatic conditions.
Farm 1 depicted a barren habitat for the pigs and was a fattening stable in which the pigs were
kept in groups of 19 to a pen. The pens measured 3.70 x 4.70m resulting in 0.92m² per pig.
The floor was half planed, and half perforated with no bedding. These pigs were fed ad
libitum with pelleted feed through a dry-feeding machine and water was accessible through
nipple drinkers. The ambient temperature was 18°C. There was daily artificial light for eight
hours (07:00-15:00).
An ecological fattening stable with an enriched habitat, such as inside and outside pens (in
total 4.90 x 9.80m) with straw bedding represented Farm 2. There was a roofed (2.70 x
21
4.90m) and roofless area (3.50 x 4,90m) in the outside pens. At the beginning of fattening 54
pigs were housed in one pen with a total area of 48.02m² (4.90 x 9.80m) resulting in 0.89m²
per pig. Throughout fattening the animals were divided into three of these pens evenly so that
each pig attained an availability of space of 2.67m². These pigs were fed ad libitum with
mealy feed.
Farm 3 constituted an ecological fattening stable with a specifically enriched habitat for the
pigs. There were straw-bedded inside pens and roofed straw-bedded outside pens as well as
outside pens with soiled floor for rooting. The dimension of these pens measured in total
83.32m² (8.33m² per pig with a usual occupancy of 10 pigs per pen). The feeding was ad
libitum with liquid feed.
On Farm 2 and Farm 3 there were also nipple drinkers and hayracks available and the daylight
length as well as the ambient temperature was determined by the season. Once a day these
pigs got bread and different fruits and vegetables.
Experimental procedure
297 fattening pigs of three different farms were subjected to three HATs and three NOTs
during their fattening. The first, second and third points of testing were conducted at the
beginning, middle and end of fattening. Each pig was tested alone in the home pen. Both
behavioural tests were performed with a one-day time lag in between and never on the same
day. The pigs were given an acclimation period of two minutes followed by a test period of
three minutes that started when the unknown human entered, or the novel object were brought
into the test area. The unknown human in the HAT was always a female person who the pigs
did not know from daily routine work. She wore a clean overall plus rubber boots and stood
motionless in the middle of the pen during the entire test phase. A plastic duck presented to
the pigs in three different sizes (according to the age and the live weight of the animals)
represented the utilized novel objects in the NOT. These ducks showed a yellow body colour
with a red coloured beak. With the help of a rod and a string, they were held in the middle of
the pen at the height of the pig's head. Always the same observing person noted the analysed
variables during the test phase: the approach latency (AL), the duration of contacts (DC) and
the number of contacts (NC). AL represented the time in seconds that each pig needed to
approach the unknown human or the novel object until the snout touched the human or the
novel object. DC exposed the accumulated seconds in which the pigs touched the human or
22
the novel object with their snouts. The entire number of snout contacts that occurred during
the test phase was indicated by NC. Further, an AL of 180 seconds was noted if the pigs did
not come into contact with the human or the novel object.
The behavioural tests were conducted in the home pens of the respective housing systems. On
Farm 1 the home pen (3.70 x 4.70m) was used completely. Parts of the outdoor area were
used for testing the pigs on the other two farms. The amount of space for testing on Farm 2
measured 6.20 x 4.90m with a roofed and unroofed area. On Farm 3, the roofed outdoor area
was utilized (2.40 x 3.80m).
Statistical analysis
Statistical analyses were performed with SAS® 9.4 (SAS Institute Inc., 2017). The
behavioural data were not normally distributed. Thus, all data were log10 (X + 1) transformed
to obtain normality of residuals of the used linear mixed model (PROC MIXED). Fixed
effects were added to the model in a stepwise manner. The Akaike’s information criteria
corrected (AICC) and the Bayesian information criteria (BIC) were used to compare the
different models. The model with the smallest AICC and BIC values was chosen and included
the fixed effects farm (1-3), batch (1, 2) nested in farm, points of testing (beginning, middle
and end of fattening) nested in farm and gender (female, male) together with a random effect
of each individual pig nested in farm, batch and gender. The significance of differences in the
least square means was adjusted with the Bonferroni-correction. Statistical significance was
determined at p<0.05.
The residuals of the linear mixed models of all behavioural variables of both behavioural tests
were correlated through the Pearson correlation coefficient (PROC CORR).
23
RESULTS
Human Approach Test (HAT)
Approach latency (AL) HAT
Differences between the farms at each point of testing
No significant differences between the AL were observed on farm 1 and farm 2 at the
beginning of fattening (Figure 1). On farm 1, the AL was significantly lower than on farm 2 at
the middle and the ending of the fattening (Figure 1). There were no significant differences
between the AL of farm 1 and farm 3 at the beginning of the fattening and significantly lower
AL on farm 1 than on farm 3 at the middle and the ending of the fattening (Figure 1). No
significant differences between the AL were observed on farm 2 and farm 3 at the beginning,
middle and ending of the fattening (Figure 1).
Differences between the points of testing within each farm
The AL significantly decreased throughout fattening from the beginning to end on farm 1 and
farm 2 (Figure 1). On farm 3, no significant differences in the AL were observed between the
beginning and end of fattening (Figure 1).
Duration of contact (DC) HAT
Differences between the farms at each point of testing
At the beginning, middle and end of fattening the pigs on farm 1 showed longer DC than the
pigs on farm 2 (Figure 2). There were no significant differences observed between the DC of
farm 1 and farm 3 at the beginning of fattening (Figure 2). At the middle and end of fattening,
the pigs on farm 1 showed longer DC than the pigs on farm 3 (Figure 2). On farm 2, the DC
were significantly lower than on farm 3 at the beginning of fattening and there were no
significant differences between the DC on farm 2 and farm 3 at the middle and end of
fattening (Figure 2).
Differences between the points of testing within each farm
On farm 1 and farm 2, the DC significantly increased from the beginning to the end of
fattening (Figure 2). On farm 3, there was no significant increase or decrease of the DC
between the beginning and end of fattening observed (Figure 2).
24
Figure 1: LS-means and standard errors (s) (retransformed) of the approach latency in the HAT at each point of testing; a, b, c different letters indicate significant differences within the farm between the different points of testing (p<0.05); A, B, C different letters indicate significant differences between the farms within the different points of testing (p<0.05).
25
Figure 2: LS-means and standard errors (s) (retransformed) of the duration of contacts in the HAT at each point of testing; a, b, c different letters indicate significant differences within the farm between the different points of testing (p<0.05); A, B, C different letters indicate significant differences between the farms within the different points of testing (p<0.05).
Number of contacts (NC) HAT
Differences between the farms at each point of testing
At the beginning and middle of fattening the pigs on farm 1 showed significantly higher NC
than the pigs on farm 2 (beginning of the fattening: farm 1: 1.8 ± 1.0s vs. farm 2 1.3 ± 1.0s;
p<0.05; middle of the fattening: farm 1: 2.9 ± 1.0s vs. farm 2: 1.8 ± 1.0s; p<0.05). At the end
of fattening there were no significant differences between the NC of farm 1 and farm 2 (farm
1: 2.3 ± 1.0s vs. farm 2: 1.9 ± 1.0s; p>0.05). There were no significant differences observed
between the NC of farm 1 and farm 3 at the beginning and end of fattening (beginning of
fattening: farm 1: 1.8 ± 1.0s vs. farm 3: 1.7 ± 1.1s; p>0.05; end of fattening: farm 1: 2.3 ± 1.0s
vs. farm 3: 1.8 ± 1.1s; p>0.05). At the middle of fattening, the pigs on farm 1 showed
significantly higher NC than the pigs on farm 3 (farm 1: 2.9 ± 1.0s vs. farm 3: 1.5 ± 1.1s;
p<0.05). On farm 2, there was significant lower NC than on farm 3 at the beginning of
26
fattening (farm 2: 1.3 ± 1.0s vs. farm 3 1.7 ± 1.1s; p<0.05) and at the middle and end of
fattening there were no significant differences between the NC of farm 2 and farm 3 (middle
of fattening: farm 2: 1.8 ± 1.0s vs. farm 3 1.5 ± 1.1s; p>0.05); end of fattening: farm 2: 1.9 ±
1.0s vs. farm 3: 1.8 ± 1.1s; p>0.05).
Differences between the points of testing within each farm
There was a significant increase in NC on farm 1 and farm 2 from the beginning to the end of
fattening (farm 1: 1.8 ± 1.0s vs. 2.3 ± 1.0s; p<0.05); farm 2: 1.3 ± 1.0s vs. 1.9 ± 1.0s; p<0.05)
and no significant differences between the NC of the beginning and end of fattening on farm 3
(1.7 ± 1.1s vs. 1.8 ± 1.1s; p>0.05).
Gender HAT
The female pigs showed significantly lower AL and longer DC than the castrated males
(Table 1). No significant differences were observed between the NC of the female and
castrated male pigs (Table 1).
Table 1: LS-means and standard errors (s) (retransformed) of the behavioural variables in the HAT and NOT of different gender
Behavioural
variables
Female Male p-value
HAT-AL 33.36 ± 1.09a 52.05 ± 1.11b 0.04
HAT-DC 10.97 ± 1.10a 7.75 ± 1.12b 0.005
HAT-NC 1.96 ± 1.03a 1.82 ± 1.03a 0.05
NOT-AL 12.52 ± 1.08a 16.14 ± 1.09b 0.01
NOT-DC 25.98 ± 1.08a 20.95 ± 1.10b 0.03
NOT-NC 3.23 ± 1.03a 3.04 ± 1.03a 0.1
HAT = human approach test; NOT = novel object test; AL = approach latency; DC = duration of contact; NC = number of contacts; a, b different letters indicate significant differences between the gender (p<0.05).
27
Novel Object Test (NOT)
Approach latency (AL) NOT
Differences between the farms at each point of testing
At the beginning, middle and end of fattening significantly lower AL were observed on farm
1 than on farm 2 (Figure 3). At the beginning and middle of fattening there were no
significant differences between the AL on farm 1 and farm 3. At the end of fattening the pigs
on farm 1 showed lower AL than the pigs on farm 3 (Figure 3). There were no significant
differences between the AL on farm 2 and farm 3 at the beginning, middle and end of
fattening (Figure 3).
Differences between the points of testing within each farm
The AL on farm 1 and farm 2 decreased from the beginning to the end of fattening (Figure 3).
No significant differences of the AL were observed at the beginning and end of fattening on
farm 3 (Figure 3).
Duration of contact (DC) NOT
Differences between the farms at each point of testing
There were longer DC on farm 1 than on farm 2 at the beginning and end of fattening and no
significant differences between the DC of farm 1 and farm 2 at the middle of fattening (Figure
4). On farm 1 and farm 3 there were no significant differences observed between the DC at
the beginning, middle and end of fattening (Figure 4). At the beginning of fattening the pigs
on farm 2 showed shorter DC than the pigs on farm 3 and there were no significant
differences between the DC of farm 2 and farm 3 at the middle and end of fattening (Figure
4).
Differences between the points of testing within each farm
The pigs on farm 1 and farm 2 showed a significant increase in DC from the beginning to the
end of fattening and there were no significant differences between the DC of the beginning
and end of fattening on farm 3 (Figure 4).
28
Figure 3: LS-means and standard errors (s) (retransformed) of the approach latency in the NOT at each point of testing; a, b, c different letters indicate significant differences within the farm between the different points of testing (p<0.05); A, B, C different letters indicate significant differences between the farms within the different points of testing (p<0.05).
29
Figure 4: LS-means and standard errors (s) (retransformed) of the duration of contacts in the NOT at each point of testing; a, b, c different letters indicate significant differences within the farm between the different points of testing (p<0.05); A, B, C different letters indicate significant differences between the farms within the different points of testing (p<0.05).
Number of contacts (NC) NOT
Differences between the farms at each point of testing
At the beginning, middle and end of fattening the pigs on farm 1 showed significantly higher
NC than the pigs on farm 2 (beginning of fattening: farm 1: 3.3 ± 1.0s vs. farm 2 1.9 ± 1.0s;
p<0.05; middle of fattening: farm 1: 3.6 ± 1.0s vs. farm 2: 2.8 ± 1.0s; p<0.05; end of
fattening: farm 1: 4.3 ± 1.0s vs. farm 2: 2.5 ± 1.0s; p<0.05). At the beginning, middle and end
of fattening there were no significant differences between the NC of farm 1 and farm 3
(beginning of fattening: farm 1: 3.3 ± 1.0s vs. farm 3 3.6 ± 1.1s; p<0.05; middle of fattening:
farm 1: 3.6 ± 1.0s vs. farm 3: 3.5 ± 1.1s; p<0.05; end of fattening: farm 1: 4.3 ± 1.0s vs. farm
3: 3.1 ± 1.1s; p<0.05). There was a significantly lower NC on farm 2 than on farm 3 at the
beginning of fattening (farm 2: 1.9 ± 1.0s vs. farm 3: 3.6 ± 1.1s; p<0.05) and no significant
differences between the NC of farm 2 and farm 3 at the middle and end of fattening (middle
30
of fattening: farm 2: 2.8 ± 1.0s vs. farm 3: 3.5 ± 1.1s; p>0.05; end of fattening: farm 2: 2.5 ±
1.0s vs. farm 3: 3.1 ± 1.1s; p>0.05).
Differences between the points of testing within each farm
On farm 1 and farm 2, there was a significant increase in NC from the beginning to the end of
fattening (farm 1: 3.3 ± 1.0s vs. 4.3 ± 1.0s; p<0.05; farm 2: 1.9 ± 1.0s vs. 2.5 ± 1.0s; p<0.05).
No significant differences were observed between the NC of the beginning and end of
fattening on farm 3 (3.6 ± 1.1s vs. 3.1 ± 1.1s; p>0.05).
Gender NOT
The female pigs showed significant lower AL and higher DC than the castrated male pigs
(Table 1). There were no significant differences observed between the NC of the female and
castrated male pigs (Table 1).
Relationship between the behavioural variables
In the HAT, the AL was negatively correlated with the NC and DC (rp = -0.62 p<0.001; rp = -
0.68 p<0.001) whereas the NC was positively correlated with the DC (rp = 0.76 p<0.001)
(Table 2).
In the NOT, there was a positive correlation observed between the NC and DC (rp = 0.74
p<0.001) and there were negative correlations between the AL and the NC respectively DC
(rp = -0.55 p<0.001 respectively rp = -0.61 p<0.001) (Table 2).
The DC of the NOT was positively correlated with the NC respectively DC of the HAT (rp =
0.21 p<0.001 respectively rp = 0.24 p<0.001) (Table 2).
31
Table 2: Pearson rank correlations between the residuals of the linear mixed models of the behavioural variables
Behavioural
variables HAT-DC HAT-AL NOT-NC NOT-DC NOT-AL
HAT-NC 0.76 -0.62 0.16 0.21 -0.16
HAT-DC -0.68 0.16 0.24 -0.19
HAT-AL -0.13 -0.18 0.18
NOT-NC 0.74 -0.55
NOT-DC -0.61
NOT-AL
Correlation coefficients in bold indicate moderate respectively high and statistically significant correlations (p<0.05).
DISCUSSION
The present study aimed to examine whether behavioural tests like the HAT or NOT
performed with fattening pigs of two different housing systems are useful to assess their
positive affective state.
In the following, the behavioural variables AL, DC and NC are discussed together for both
behavioural tests. The AL indicated definite differences in the pigs’ behaviour between the
two different housing systems in both the HAT and NOT especially at the middle and end of
fattening. The pigs housed in the barren environment showed lower AL than the enriched-
housed pigs. The same pattern of behaviour was observed for the DC in the HAT but not in
the NOT. Finally, the NC showed no definite differences of the pigs’ behaviour between the
two housing systems, neither in the HAT nor in the NOT.
Approach latency (AL)
According to Brown et al. (2009) quicker latencies to approach novel stimuli, such as
unknown humans or novel objects, are associated with less fearful animals. Therefore it could
be imaginable that these barren-housed pigs, which showed significantly quicker AL, possess
a more positive affective state due to less fearfulness but the quicker AL could also be a
reason for a stronger motivation to explore novel stimuli (Stolba and Wood-Gush, 1980)
32
related to boredom in the environment of these barren-housed pigs. Additionally,
Wemelsfelder et al. (2000) argued that enriched-housed pigs might be less fearful or more
curious to explore a novel object or a person in the home pen. But, simultaneously, barren-
housed pigs seemed to be more motivated to explore a novel object or a person in the home
pen probably due to fewer exploration possibilities in their habitat compared to enriched-
housed pigs. Further on, Forkman et al. (2007) claim that the avoidance reaction of the animal
appears to be essential as both a non-curious and a fearful animal will show long latencies to
approach. Consequently, it is difficult to determine, whether the quicker AL of the barren-
housed pigs might represent behavioural indicators for a more positive affective state possibly
because of less fearfulness, because other authors maintain that barren environments have
negative effects on pig welfare (Beattie et al., 1995). Thus, it could be also conceivable that
these quicker AL of the pigs housed in the barren habitat describe their more negative
affective state, possibly due to boredom and fewer exploration opportunities in their
environment.
The decreasing AL from the first to the third points of testing during fattening on farm 1 and
farm 2 both in the HAT and NOT could be related to a gradual reduction in fear;
Wemelsfelder et al. (2000) also observed a decreasing latency to enter the test arena for all
pigs while the experiment progressed. Additionally, Forkman et al. (2007) suggest that fear
and anxiety decrease with age, which could also explain the decreasing AL during fattening
on farm 1 and farm 2. The reduced level of fear and anxiety probably also causes the pigs to
get used to unknown humans or novel objects (Forkman et al., 2007) so that AL decreases
when the fattening progresses.
Moreover, an intensive monitoring of the avoidance reaction of the pigs e.g. regarding a high
latency to approach the unknown human or the novel object might be able to evaluate whether
the pigs show avoidance reactions or do not approach the novel stimuli due to lack of interest
(possibly related to various exploration opportunities in the environment) or fear and anxiety
accompanied by flight behaviour. The implementation of a forced human approach test might
be a possible solution to detect the reason of the avoidance reaction of the pigs whereby
animals which flee when humans approach them are known to be fearful (Pedersen et al.,
2003). Also Waiblinger et al. (2006) suggest that forced and voluntary approach tests measure
the animals’ behaviour in differing ways: the forced approach test might increase the
likelihood that an animal responds more actively to the human whereas in the voluntary
33
approach test the chances of getting no response or a passive response might probably be
higher. Hence, related observations between high approach latencies and the avoidance
reaction could provide more detailed information about a more positive, respectively more
negative affective state of the examined pigs.
Duration of contacts (DC)
In the HAT, there was a longer DC observed in the barren housing system (farm 1) than in the
enriched housing system (farm 2 and farm 3) at the middle and end of fattening. This definite
difference between the DC of the pigs in the two different housing systems could also be
related to the level of motivation to explore novel stimuli. Stolba and Woodgush (1981) argue
that the pigs’ interaction with a novel object decreased with increasing environmental
complexity and that the pigs in more intensive housing systems were more interested in the
novel object and that their interest maintained for a longer period. Consequently, this longer
DC related to a high level of motivation to explore novel stimuli in the barren-housed pigs
could imply a behavioural indicator of less welfare and therefore a less positive affective state
of these animals. Nevertheless, this definite difference in the pigs’ behaviour between the two
housing systems was only observed in the HAT. In the NOT, there were significant
differences of the pigs’ behaviour between the three farms, but not between the two housing
systems. Hence, this might imply that the HAT and NOT do not measure the same
behavioural attributes. Also Boivin et al. (1992) found no relationship between open-field
tests and handling tests, possibly indicating that they do not reflect the same animal
characteristic, which was also displayed in the weak correlations between the behavioural
variables of the HAT and NOT. For example, in contrast to the NOT, the pigs’ behaviour in
the HAT was influenced by their previous experiences with humans. There are studies which
demonstrate that negative handling leads to more avoidance responses (Carreras et al., 2017).
Increasing DC was observed on farm 1 and farm 2 from the first to the third points of testing
during the fattening both in the HAT and NOT. Regarding this, Forkman et al. (2007) posit
that human and object investigation increases with age, suggesting that fear and anxiety
decrease with age and that there could be an habituation process that reduces the fear
responses.
34
Number of contacts (NC)
At the second point of testing in the HAT and third point of testing in the NOT, higher NC
was observed in the barren housing system (farm 1) than in the enriched housing system (farm
2 and farm 3). Reimert et al. (2014) also observed barren-housed pigs more frequently present
near the person than enriched-housed pigs. This can also be justified by less fearful pigs
respectively a higher motivation for exploring novel stimuli but again it is not clear whether
this pattern of behaviour is related to a more positive or more negative affective state.
Besides, this definite difference in the pigs’ behaviour was only shown at one of the six points
of testing during fattening in the HAT respectively NOT.
The increased NC from the first to the third point of testing during fattening on farm 1 and
farm 2 in the HAT and NOT might also be related to a habituation process (Forkman et al.,
2007), which lowered the level of fear and thereby increased the NC of the pigs with the
unknown human or the novel object.
Gender
The female pigs showed quicker AL and longer DC than the male pigs in both, the HAT and
NOT. This might indicate that female pigs are less fearful than male pigs. Reimert et al.
(2014) observed gilts approaching and touching a person and a novel object faster than the
barrows suggesting the female animals were less fearful than the male ones. The gilts of this
study also had lower basal cortisol concentrations than barrows (Reimert et al., 2014), which
could explain the reaction of the female pigs as less fearful since high cortisol concentrations
depicted an indication of stress (Jarvis et al., 2006), and stress is known to be often connected
with fear (Forkman et al., 2007).
Further, especially the higher AL of the male pigs than the female pigs in the HAT can be
explained by their castration and its consequences, since in studies by Reimert et al. (2013)
the castrated male piglets also responded more fearfully to novel situations than female
piglets. Additionally, in this study, the male pigs had higher salivary cortisol concentrations
than the female pigs (Reimert et al., 2013), which could also indicate a stressed and fearful
affective state, as mentioned before. It is supposed that the absence of gonadal testosterone
makes the castrated pigs more fearful (Reimert et al., 2013) and that handling by a human
during castration and the connected pain afterwards makes the castrated male pigs more
fearful of humans (Prunier et al., 2006) so that they approach them not as quickly as the
35
female pigs. Further studies with uncastrated boars would be useful to support these
assumptions.
Additionally, the development of brains could influence the exploratory behaviour of animals;
Fleming and Dilger (2017) claim that, in general, the female brain of pigs develops faster than
the male ones possibly explaining the shorter AL and longer DC of the female pigs in both
behavioural tests due to better developed cognitive abilities.
CONCLUSION
The results of the present study showed that HAT and NOT performed with fattening pigs of
two different housing systems might be suitable to assess their level of anxiety and fear, but,
simultaneously, they also appeared to be capable of showing a stronger or lower motivation to
explore novel stimuli making it difficult to draw a clear conclusion as to whether a high or
low level of motivation to explore indicates more negative or more positive affective states.
Conclusively, in this study, neither the HAT nor the NOT depicted autonomous, reliable
indicators of positive affective states in growing pigs. However, the results of this study lay
essential foundations for further investigations concerning the assessment of the positive
affective state of fattening pigs. Furthermore, the findings of this study provide more detailed
information about certain reactions to unknown humans or novel objects of fattening pigs,
which are housed in different environments.
ACKNOWLEDGEMENTS
The project is supported by funds of the Federal Ministry of Food and Agriculture (BMEL)
based on a decision of the Parliament of the Federal Republic of Germany via the Federal
Office for Agriculture and Food (BLE) under the innovation support programme.
36
REFERENCES
Beattie, V.E., Walker, N., Sneddon, I.A., 1995. Effects of environmental enrichment on
behaviour and productivity of growing pigs. Animal Welfare 4, 207-220.
Boissy, A., 1995. Fear and fearfulness in animals. Quarterly Review of Biology 70, 165-191.
Boivin, X., Le Neindre, P., Chupin, J.M., 1992. Establishment of cattle–human relationships.
Applied Animal Behaviour Science 32, P. 325–335.
Bokkers, E.A.M., de Vries, M., Antonissen, I., de Boer, I.J.M., 2012. Inter- and intra-observer
reliability of experienced and inexperienced observers for the Qualitative Behaviour
Assessment in dairy cattle. Animal Welfare 21, 307-318.
Bracke, M., Spoolder, H.A.M., 2008. Novel object test can detect marginal differences in
environmental enrichment in pigs. Applied Animal Behaviour Science 109 (1), 39–48.
Brown, J.A., Dewey, C., Delange, C.F.M., Mandell, I.B., Purslow, P.P., Robinson, J.A.,
Squires, E.J., Widowski, T.M., 2009. Reliability of temperament tests on finishing
pigs in group-housing and comparison to social tests. Applied Animal Behaviour
Science 118, 28-35.
Carreras, R, Arroyo, L., Mainau, E., Valent, D., Bassols, A., Dalmau, A., 2017. Can the way
pigs are handled alter behavioural and physiological measures of affective state?
Behavioural processes 142, 91–98.
Carter, A. J., Marshall, H. H., Heinsohn, R., Cowlishaw, G., 2012. How not to measure
boldness. Novel object and antipredator responses are not the same in wild baboons.
Animal Behaviour 84 (3), 603–609.
Czycholl, I., Kniese, C., Schrader, L., Krieter, J., 2017. Assessment of the multi-criteria
evaluation system of the Welfare Quality® protocol for growing pigs. Animal. 11,
1573–1580.
Dalmau, A., Fabrega, E., Velarde, A., 2009. Fear assessment in pigs exposed to a novel object
test. Applied Animal Behaviour Science 117 (3-4), 173–180.
Fleming, S. A., Dilger, R. N., 2017. Young pigs exhibit differential exploratory behavior
during novelty preference tasks in response to age, sex, and delay. Behavioural brain
research 321, 50–60.
37
Forkman, B., Boissy, A., Meunier-Salaün, M.-C., Canali, E., Jones, R. B., 2007. A critical
review of fear tests used on cattle, pigs, sheep, poultry and horses. Physiology &
Behavior 92 (3), 340–374.
Fraser, D., 2009. Animal behaviour, animal welfare and the scientific study of affect. Applied
Animal Behaviour Science 118, 108-117.
Hemsworth, P.H., Coleman, G.J., 1998. Human–Livestock Interactions: The Stockperson and
the Productivity of Intensively Farmed Animals. CAB International, Wallingford, UK.
Jarvis, S., D’Eath, R.B., Robson, S.K., Lawrence, A.B., 2006. The effect of confinement
during lactation on the hypothalamic-pituitary-adrenal axis and behaviour of
primiparous sows. Physiology and Behaviour 87, 345-352.
Marcet Rius, M., Cozzi, A., Bienboire-Frosini, C., Teruel, E., Chabaud, C., Monneret, P.,
2018. Selection of putative indicators of positive emotions triggered by object and
social play in mini-pigs. Applied Animal Behaviour Science 202, 13–19.
Murphy, E., Nordquist, R.E., van der Staay, F.J., 2014. A review of behavioural methods to
study emotion and mood in pigs, Sus scrofa. Applied Animal Behaviour Science. P.
9-28.
Pedersen, L.J., Damm, B.I., Kongsted, A.G., 2003. The influence of adverse or gentle
handling procedures on sexual behaviour in fearful and confident sows. Applied
Animal Behaviour Science 83, 277-290.
Prunier, A., Bonneau, M., von Borell, E.H., Cinotti, S., Gunn, M., Frederiksen, B., Giersing,
M., Morton, D.B., Tuyttens, F.A.M., Velarde, A., 2006. A review of the welfare
consequences of surgical castration in piglets and the evaluation of non-surgical
methods. Animal Welfare 15, 277-289.
Reimert, I., Rodenburg, T. B., Ursinus, W. W., Kemp, B.; Bolhuis, J. E., 2014. Responses to
novel situations of female and castrated male pigs with divergent social breeding
values and different backtest classifications in barren and straw-enriched housing.
Applied Animal Behaviour Science 151, 24–35.
Reimert, I., Rodenburg, T.B., Ursinus, W.W., Duijvesteijn, N., Camerlink, I., Kemp, B.,
Bolhuis, J.E., 2013. Backtest and novelty behavior of female and castrated male
piglets with diverging social breeding values for growth. Journal of Animal Science
91, 4589-4597.
38
SAS (2017): SAS Institute Inc. Cary, NC, USA.
Stolba, A., Wood-Gush, D.G.M., 1980. Arousal and exploration in growing pigs in different
environments. Applied Animal Ethology, 382-383.
Stolba, A., Woodgush, D.G.M., 1981. Verhaltensgliederung und Reaktion auf Neureize als
ethologische Kriterien zur Beurteilung von Haltungsbedingungen bei Hausschweinen.
KTBL-Schrift 264, 110-128.
Temple, D., Dalmau, A., Ruiz de la Torre, J., Manteca, X., Velarde, A., 2011. Application of
the Welfare Quality protocol to assess growing pigs kept under intensive conditions in
Spain. Journal of Veterinary Behavior 6:138–149.
Tuyttens, F.A.M., de Graaf, S., Heerkens, J.L.T., Jacobs, L., Nalon, E., Ott, S., Stadig, L.,
Van Laer, E., Ampe, B., 2014. Oberserver bias in animal behaviour research: can we
believe what we score, if we score what we believe? Animal Behaviour 90, 273-280.
van der Staay, F. J., van Zutphen, J.A., Ridder, M. M. de; Nordquist, R. E., 2017. Effects of
environmental enrichment on decision-making behavior in pigs. Applied Animal
Behaviour Science 194, 14–23.
Waiblinger, S. Boivin, S., Pedersen, V., Tosi, M., Janczak, A.M., Visser, E.K., 2006.
Assessing the human animal relationship in farmed species: a critical review. Applied
Animal Behaviour Science 101, 185-242.
Wemelsfelder, F., Haskell, M., Mendl, M.T., Calvert, S., Lawrence, A.B., 2000. Diversity of
behaviour during novel object tests is reduced in pigs housed in substrate-
impoverished conditions. Animal Behaviour 60 (3), 385–394.
Wilsson, E. and Sundgren, P.E., 1997. The use of a behaviour test for selection of dogs for
service and breeding. II. Heritability for tested parameters and effect of selection based
on service dog characteristics. Applied Animal Behaviour Science 54 (2-3), 235-241.
39
CHAPTER TWO
Playing behaviour – a suitable indicator to measure positive emotions in
growing pigs?
K.L. Krugmann , F.J. Warnken, J. Krieter, I. Czycholl
Institute of Animal Breeding and Husbandry, Christian-Albrechts-University Kiel,
Olshausenstr. 40, D-24098 Kiel, Germany.
Submitted to Journal of Animal Science
40
ABSTRACT
There is a lack of suitable measurement methods which are established for determining the
well-being of livestock. Thus, the aim of the present study was to examine whether the
occurrence of playing behaviour in fattening pigs (n=228) in different housing systems can be
used as an indicator to identify their positive affective state. The investigated housings
consisted of a barren (farm 1, n=138) and two enriched (farm 2, n=54; farm 3, n=36) habitats
that varied in terms of availability of space, enrichment and seasonal influences. The
behaviour of the pigs was recorded for two days at the beginning of fattening (first fattening
phase) and two days at the end of fattening (final fattening). In the first fattening phase, the
pigs in the barren habitat played significantly more than those in the enriched habitats (farm 1:
8.3 ± 1.1 s/h; p<0.05 vs. farm 2: 1.5 ± 1.0 s/h; farm 3: 3.3 ± 1.0 s/h). In the final fattening
phase, this pattern was reversed with pigs showing significantly more playing behaviour in
the enriched environment (farm 2: 2.2 ± 1.0 s/h; farm 3: 2.7 ± 1.0 s/h vs. farm 1: 1.2 ± 1.0 s/h;
p<0.05). The locomotor playing behaviour did not indicate definite differences between the
two housing systems in the first fattening phase, but significantly more locomotor play in the
enriched than barren housing system in the final fattening (final fattening: farm 1: 0.7 ± 1.0
s/h vs. farm 2: 1.6 ± 1.0 s/h; farm 3: 1.3 ± 1.0 s/h; p<0.05). Conversely, the pigs in the barren
habitat showed significantly more social playing behaviour than the pigs in the enriched
housing system in the first fattening phase and there were no definite differences between the
two housing systems in the final fattening (first fattening phase: farm 1: 6.6 ± 1.0 (s/h) vs.
farm 2: 1.2 ± 1.0 s/h; farm 3: 2.0 ± 1.0 s/h; p<0.05). Taken together, these observations
clearly show that the occurrence of playing behaviour in fattening pigs varies across fattening
phases and different housing systems. However, it remains questionable and difficult to
interpret whether playing behaviour as a solitary indicator is reliable to determine positive
affective states in fattening pigs.
Keywords: affective state, animal welfare, pigs, playing behaviour, positive emotions
INTRODUCTION
Modern livestock farming is currently intensively involved in public and political debate.
Consumers are concerned about the well-being of farm animals, housing conditions have been
41
criticised and an increased focus on animal welfare is required (Rushen, 2003; Veissier et al.,
2012; Alban et al. 2015). Consequently, current science includes the aim of researching the
objective detection of livestock’s well-being (Webster, 2005). It has already been defined that
animal welfare consists of three basic parts: basic health and functioning, natural living, and
affective state (Fraser, 2008). Whereas extensive research in the area of the basic health and
functioning and natural living has already been carried out (Blokhuis, 2013), there is a need to
detect the affective state of farm animals more intensively, which represents the pleasant or
unpleasant emotions they experience (Fraser, 2009). It has also been established that animals
are able to be aware to experience basic emotions such as happiness or anxiety (Panksepp,
1998). Up to now previous assessment systems like e.g. the “Welfare Quality®” protocol has
been partially unreliable and strongly influenced by subjective perceptions, especially in
terms of the “Qualitative Behaviour Assessment” (Czycholl et al., 2017; Temple et al., 2011).
Hence, the measurement of affective state is difficult and an enormous challenge (Duncan,
2005). Accordingly, it is further necessary to identify reliable and suitable indicators which
detect a livestock’s affective state and especially their positive emotions (Marcet Rius et al.,
2018).
Playing behaviour occurs in mammals, can be measured non-invasively and is easy to
recognise (Fraser and Duncan, 1998; Špinka et al., 2001; Barnard, 2004). It represents not
completely functional behaviour and does not contribute to the survival of animals
(Burghardt, 2005) as e.g. feed or water intake. Hence, playing behaviour includes apparent
main types of social (Vanderschuren et al., 1997), locomotor-rotational (Newberry et al.,
1988) and object play (Bateson and Young, 1981). Further, it has been stated that playing
only occurs when optimal environmental conditions prevail (Lawrence, 1987; Held and
Špinka, 2011) and the primary needs of animals (for food, safety, comfort, etc.) have reached
a satisfactory level (Buchenauer, 1981; Newberry, 1988). Authors have also suggested that
playing behaviour exists if the animals fare well and feel well (Siviy et al., 2006) especially in
terms of being free from sickness, injury, thermal stress or other challenges to their fitness and
health (Martin and Caro 1985; Burghardt, 2005). Moreover, it increases with improved
physical condition (Sharpe et al., 2002; Nunes et al., 2004; Cameron et al., 2008). Thus,
playing behaviour in livestock can be seen as “luxury” behaviour (Lawrence, 1987), which
decreases when animals experience negative emotions accompanied by e.g. threats to their
fitness (Fagen 1976; Martin and Caro 1985; Burghardt 2005) or adverse environmental
conditions (Müller-Schwartze et al., 1982; Siviy and Panksepp, 1985). Additionally, it has
42
been said to be often connected with animals experiencing positive emotions (Fraser and
Duncan, 1998; Špinka et al., 2001; Barnard, 2004; Burgdorf and Panksepp, 2006).
Furthermore, the playing behaviour of farm animals has been identified as a potential
indicator of current well-being (Fagen, 1976; Lawrence, 1987) and according to Boissy et al.
(2007), playing behaviour represents one of the most important indicators of positive
emotions in captive animals. Domestic pigs (Sus scrofa domesticus) are considered social,
intelligent mammals (D’Eath and Turner, 2009; Gieling et al., 2011) and research shows that
pig playing behaviour occurs beyond all playing behaviour categories: locomotor, social and
object (Donaldson et al., 2002; Newberry et al., 1988).
Along with the research aim of current scientific study this paper examined whether the
playing behaviour of fattening pigs in different housing systems can be used as an objective
measurement method to assess affective state. It has been hypothesised that the occurrence of
more playing behaviour in pigs might indicate their more positive affective state.
MATERIALS AND METHODS
Animals and housing
The data were collected on three farms with different housing systems in Northern Germany
from November 2016 until September 2017. 228 fattening pigs bred from commercial cross-
bred dams (Large White x Landrace) and sired by Pietrain boars were used in total. All boars
were castrated and the tails were undocked. The housing systems differed especially in terms
of an enriched or barren habitat, the availability of space and climatic conditions. The pigs in
the barren housing system were weaned at the age of four weeks and those of the enriched
farms were weaned at six weeks.
Farm 1 depicted a conventional fattening stable (n=138) managed in a closed system in which
the pigs were housed in groups of 19 to a pen. These pens measured 3.70 x 4.70m resulting in
0.92m² per pig and the floor was half perforated and half concrete. Thus, farm 1 represents the
barren environment. The pigs were fed ad libitum with pelleted feed through a dry-feeding
machine involving an animal feeding space ratio of 1:4. Water was accessible through nipple
drinkers and the ambient temperature was 18° C. Metal chains, plastic pipes and balls were
43
offered as occupation material and in case of tail biting, the pigs were given gunny sacks.
Artificial light was provided for eight hours (07:00-15:00) per day.
Farm 2 (n=54) represented an ecological fattening stable with an enriched habitat, e.g. inside
and outside pens (in total 4.90 x 9.80m) with straw bedding. The outside pen included a
roofed (2.70 x 4.90m) and roofless area (3.50 x 4,90m). In the first fattening phase, there were
54 pigs housed in one pen with a total area of 48.02m² (4.90 x 9.80m) giving 0.89m² per pig.
After the first fattening phase, the animals were evenly divided into three of these pens so that
each pig had 2.67m² available space. The feeding was ad libitum with mealy feed by an
animal feeding space ratio of 1:4.
Also, the third farm (farm 3; n=36) constituted an ecological fattening stable with an enriched
environment. There were straw-bedded inside pens, roofed straw-bedded outside pens and
additionally special soil-based rooting areas. The whole pens measured 83.32m² in total
(8.33m² per pig with a usual occupancy of 10 animals per pen). The pigs were fed ad libitum
with liquid feeding including an animal feeding space ratio of 1:1.
On farms 2 and 3 there were also hay racks and nipple drinkers available but no occupation
material like metal chains, plastic pipes or balls as on farm 1. The ambient temperature as
well as the daylight length was determined by the season. Additionally, these pigs received
bread and different vegetables and fruits daily.
Experimental design
This study examined the playing behaviour of fattening pigs at the first fattening phase and at
the final fattening of two batches. The playing behaviour was determined by an ethogram
(Table 1) basing on previous research.
Recording and analysis of playing behaviour
The behaviour of the pigs was recorded for two days at the beginning and two days at the end
of the fattening using camera systems (HeiTel Digital Video GmbH, Kiel, Germany and
AXIS M30-VE Network Cameras). The cameras were positioned above the pens to obtain a
complete overview. Each of the pigs was marked individually with a sign on its back.
Continuous sampling was used to note the duration (s/h) of each playing behaviour sequence
(Table 1). Due to the great extent of the video data, it was assessed by four observers. These
observers were intensively trained using video test sequences at the beginning of the video
44
analysis and through repeated courses of instruction between the individual evaluation
sections. Additionally, the observers were applied cross-classified and not replaced during the
whole video analysis process. If several pigs played simultaneously, the sequence was
analysed for one pig and then rewound and analysed for the others. According to the testing
environment, the collected video material was examined following the different play
categories (Table 1) for eight hours per day for farm 1 and corresponding to the season for the
daylight hours for farms 2 and 3. Thus, the total duration of playing behaviour was divided by
the light day per hour in order to obtain a comparable attribute: the total duration of playing
behaviour (s/h). This procedure was also performed to obtain the total durations of playing
behaviour (s/h) for each play category (Table 1). The total duration of the locomotor play
(s/h) included all play categories that were shown exclusively by body movements whereas
the total duration of the social play (s/h) indicates all play categories with additional social
issues (Table 1).
Table 1: Ethogram of pigs’ playing behaviour
Play category Code Description References
Locomotor play Pivot a Gambolling or twirling the body by
1 to 360° Brown et al. (2015), Chaloupkova et al. (2007), Donaldson et al. (2002), Newberry et al. (1988)
Hop b Jumping off with the hind legs and angling the forelegs
Welker (1961)
Scamper c Energetic running (from 1m or along the whole pen)
Brown et al. (2015)
Flop d Dropping from an upright to a lying or sitting position on the pen floor
Brown et al. (2015), Chaloupkova et al. (2007), Donaldson et al. (2002)
Roll e Rolling lengthwise over the back Defined for this study Head-shaking j Moving the head back and forth
quickly with ears flapping Reimert et al. (2013)
Social play Non-harmful
fighting g Heads and mouths playfully
touching each other (no pushing anti-parallel/ without apparent aggression)
Brown et al. (2015)
Chase h Chasing other pigs playfully (without apparent aggression)
Welker (1961)
Invitation i Pushing the snout with minimal or moderate force into another pig’s body (if playing resulted afterwards)
Martin et al. (2015)
45
Statistical analysis
Statistical analyses were performed with SAS® 9.4 (SAS Institute Inc., 2017). The
behavioural data were not normally distributed. Thus, all data were log10 (X + 1) transformed
to obtain normality of residuals of the used linear mixed models (PROC MIXED). Fixed
effects were added to the model in a stepwise manner. The Akaike’s information criteria
corrected (AICC) and the Bayesian information criteria (BIC) were used to compare the
different models and the model with the smallest AICC and BIC values was chosen. The
model included the fixed effects observer (1-4), farm (farm 1, farm 2 and farm 3), gender
(female, male), points of testing (first fattening phase, final fattening) nested in farm, day (1,
2) nested in farm together with a random effect of each individual pig nested in farm and
gender and was used to analyse the data of the total duration of playing behaviour (s/h) and
total duration of locomotor respectively social play (s/h). The same model with the additional
fixed effect location nested in farm was utilised to compare the total duration of playing
behaviour (s/h) between the inside and outside area of farms 2 and 3. At this, a data-set was
used that contained only these two farms. Statistical significance was determined at p<0.05.
The significance of differences in the least square means was adjusted with the Bonferroni-
correction.
The differences between the farms within the individual play categories were analysed
descriptively using the Kruskal-Wallis test as there were too few observations in some play
categories to apply statistical models.
RESULTS
The following section represents first the results of each play category, i.e. how many pigs
performed which play category per farm. Then the retransformed LS-Means of the total
duration of playing behaviour (s/h) (all play categories summarised), of the locomotor play
(s/h) (all locomotor play categories summarised) and of the social play (s/h) (all social play
categories summarised) are presented.
Play categories
As illustrated in Figure 1, in the first fattening phase, the pigs on farm 1 performed no playing
behaviour in the categories “pivoting” (a), “hopping” (b) and “rolling” (e) whereas in the final
46
fattening these pigs showed no “rolling” (e) and no “chasing” (h) (Figure 1). Each play
category was observed during both phases on farm 2, except the play category “rolling” (e) in
the final fattening. On farm 3, each play category was observed during both phases.
Figure 1: Playing pigs (%) per farm within each play category. 1.=first fattening phase; 2.=final fattening; a=pivoting; b=hopping; c=scampering; d=flopping; e=rolling; g=non-harmful fighting; h=chasing; i=invitation play; j=head-shaking. a, b, c different letters indicate significant differences between the farms within the different points of testing (p<0.05)
Total duration of playing behaviour (s/h)
Regarding the comparison between the farms at each point of testing, the pigs on farm 1
played significantly longer than the pigs on farm 2 and farm 3 in the first fattening phase
(Figure 2). In the final fattening, significantly shorter durations of playing behaviour (s/h)
occurred on farm 1 compared to farm 2 and farm 3 (Figure 2).
Concerning the comparison between the points of testing within each farm, the pigs on farm 1
showed significantly more playing behaviour in the first fattening phase than in the final
fattening (Figure 2). The pigs on farm 2 played significantly shorter in the first fattening
47
phase than in the final fattening (Figure 2). On farm 3, there was no significant difference
observed between the first fattening phase and the final fattening (Figure 2).
In terms of the comparison between the genders, there were significantly longer durations of
playing behaviour (s/h) observed in the sows than in the castrates (sows: 2.8 ± 1.0 s/h vs.
castrates: 2.4 ± 1.0 s/h; p<0.05).
Relating to the comparison between the location (inside vs. outside area) on farm 2 and farm 3
the pigs showed significantly longer durations of playing behaviour (s/h) in the outside areas
than in the inside areas (outside: farm 2: 3.6 ± 1.0 s/h; farm 3: 4.4 ± 1.0 s/h vs. inside: farm 2:
1.2 ± 1.0 s/h; farm 3: 2.2 ± 1.0 s/h; p<0.05).
Figure 2: LS-Means and standard errors of the total duration of playing behaviour (s/h) during the fattening; a, b, c different letters indicate significant differences within the farm between the different points of testing (p<0.05); A, B, C different letters indicate significant differences between the farms within the different points of testing (p<0.05); 1.=First fattening phase; 2.=Final fattening.
48
Locomotor play (s/h)
In respect of the comparison between the farms at each point of testing, the pigs on farm 1
showed significantly more locomotor play (s/h) than the pigs on farm 2 in the first fattening
phase and there were no significant differences observed between the locomotor play (s/h) of
farm 1 and farm 3. The pigs on farm 2 showed less locomotor play (s/h) than the pigs on farm
3. In the final fattening, there occurred significantly shorter durations of locomotor play (s/h)
on farm 1 than on farm 2 and farm 3 (Table 2).
Concerning the comparison between the points of testing within each farm, on farm 1 the pigs
showed significantly more locomotor play in the first fattening phase than in the final
fattening. On farm 2, there were no significant differences between the total durations of
locomotor play (s/h) in the first fattening phase and in the final fattening. The pigs on farm 3
showed significantly more locomotor play in the first fattening phase than in the final
fattening (Table 2).
The comparison of gender of the total duration of locomotor play (s/h) showed longer
durations of locomotor play (s/h) in sows than in castrates, although this difference did not
reach significance (sows: 1.5 ± 1.0 s/h vs. castrates: 1.4 ± 1.0 s/h; p=0.06).
Table 2: LS–Means and standard errors (retransformed) of the total duration of locomotor play (s/h) and social play (s/h)
Point of testing Play
category
Farm 1
(n=138)
Farm 2
(n=54)
Farm 3
(n=36)
First fattening phase locomotor 2.27 ± 1.07A,a 1.37 ± 1.05a,B 2.04 ± 1.05A,a
Final fattening locomotor 0.79 ± 1.05A,b 1.67 ± 1.05a,B 1.39 ± 1.05B,b
First fattening phase social 6.64 ± 1.09A,a 1.24 ± 1.07a,B 2.09 ± 1.07a,C
Final fattening social 1.33 ± 1.07A,b 1.47 ± 1.08A,a 2.18 ± 1.07a,B
A, B, C different letters indicate significant differences between the farms within each point of testing (p<0.05); a, b different letters indicate significant differences within the farms between the different points of testing (p<0.05).
Social play (s/h)
Regarding the comparison between the farms at each point of testing, the pigs on farm 1
showed significantly longer durations of social play (s/h) than the pigs on farm 2 and farm 3
in the first fattening phase. Besides, there was less social play (s/h) observed on farm 2 than
on farm 3 (Table 2). In the final fattening, there were no significant differences of the social
49
play (s/h) of farm 1 and farm 2 whereby the pigs on farm 1 showed significantly less social
play (s/h) than the pigs on farm 3. On farm 2, there was significantly less social play (s/h)
than on farm 3 (Table 2).
In terms of the comparison between the points of testing within each farm, the pigs on farm 1
showed significantly longer durations of social play (s/h) in the first fattening phase than in
the final fattening. The comparison of social play (s/h) between the first fattening phase and
the final fattening presented no significant differences on farm 2 and farm 3 (Table 2).
Concerning the comparison between the genders, the sows indicated significantly longer
durations of social play (s/h) than the castrates (sows: 2.1 ± 1.0 s/h vs. castrates: 1.9 ± 1.0 s/h;
p<0.05).
DISCUSSION
The main aim of this study was to investigate whether the occurrence of playing behaviour in
fattening pigs in different housing systems is suitable to measure their positive emotions and
to obtain a better general understanding of their whole affective state.
Play categories
The pigs in the barren environment showed less variety in their playing behaviour than the
pigs in the enriched housing systems: they showed no play categories such as “pivoting”,
“hopping” and “rolling” in the first fattening phase and no “rolling” and no “chasing” in the
final fattening. These lack of variety in playing behaviour are based on Newberry et al.
(1988), who argue that pigs in intensive housing conditions are unable to express their full
repertoire of playful behaviour due to e.g. limitations of space and novel objects. Following
the hypothesis that playing behaviour occurs when animals are in positive emotional states
(Burghardt, 2005), these findings probably indicate fewer positive emotions in pigs housed in
such impoverished conditions whose playing behaviour is presumably hampered.
Total duration of playing behaviour
In the first fattening phase it is conspicuous that the pigs in the barren environment showed
considerably more playing behaviour than the pigs in the enriched housing systems. This may
be due to the different management of the two housing systems: farm 1 represented a closed
50
system, in which, the piglets changed their compartment and pens within the farm and not
through stressful transport by trucks as applied to farm 2 and farm 3. In contrast to the young
pigs in the enriched housing systems, the barren housed pigs did not have to get used to new
farmers, feeding systems and daily routines either. This holistic familiarisation with the new
situation might become a major cause of stress for the relocating pigs. Therefore, recently
experienced stress could be a reason for the lower occurrence of playing behaviour in the
enriched housing systems in the first fattening phase (Bateson, 2014). The fact that the rearing
phase was shorter in the enriched housed pigs than in the barren housed pigs might also
explain their lower occurrence of playing behaviour (s/h) at this level of age because,
according to several studies, weaning results in the decline of playing behaviour (e.g.
Donaldson et al., 2002). Presumably, these enriched-habitat pigs also suffered a relative
deterioration of their familiar habitat due to a less attractive stable system compared to the
piglet production farms where they were reared in extensive pasture conditions. This possibly
resulted in frustrated fattening pigs that did not feel well and thus performed less playing
behaviour (Le Floc'h et al., 2010; Siviy et al., 2006). Frustration also counts as a clear
indicator of poor welfare (Broom, 1991). According to this, it is conceivable that the absence
or lower occurrence of playing behaviour represents an affective state of the examined pigs,
which includes a lower number of positive emotions.
In the final fattening, the pigs in the barren environment (farm 1) showed significantly less
playing behaviour (s/h) than those in the enriched environment (farm 2 and farm 3).
Particularly on the basis of these results, it is conceivable that play stimulating circumstances
such as food enrichment, or sudden impressions such as windy weather or novel events
(Newberry et al., 1988), or the offering of straw-bedding or elements of unpredictability to the
environment (Špinka et al., 2001) dominate more likely in enriched housing systems such as
on farm 2 and farm 3. Based on these results, it could be assumed that the lower occurrence of
playing pigs in the barren environment (farm 1) reveals their affective states, which appear to
possess fewer positive emotions than those pigs in the enriched housing systems (farm 2 and
farm 3). Mainly the increased availability of space could explain the longer occurrence of
playing pigs in the enriched housing systems (farm 2 and farm 3) compared to the barren
environment (farm 1). The enriched housing systems offer the three respectively eight times
as much space availability for the pigs. These findings correspond with the view of Lawrence
(1987) that playing behaviour occurs when the environmental conditions are optimal.
Furthermore, there was no tail biting observed in the enriched housing systems, whereas this
51
behavioural disorder increasingly occurred in the barren environment (20% of the pigs were
affected by tail-biting (biter or victim)) especially during the final fattening that possibly
points to poor welfare (Brunberg, 2011) and could explain the lower occurrence of playing
behaviour of these pigs at this level of age.
Moreover, the proposal of enrichment such as straw-bedding and different vegetables and
fruits once a day might be reasonable for the increased playing behaviour in the pigs in the
enriched housing systems (farm 2 and farm 3) at the final fattening compared to the barren
environment (farm 1). Enrichment such as straw-bedding or special soil-based rooting areas
also provide opportunities for pigs to perform their natural living habits, which indicates an
important part of good welfare (Fraser, 2008) and could be assumed as an explanation for
enhanced playing in the enriched housed pigs. Vinke et al. (2005) observed juvenile mink
with access to a swimming pool performing more play behaviour, although it occurred outside
the pool. Also, pigs showed longer durations of playing, after receiving food enrichment such
as seeds scattered in straw and buried chocolate raisins (Dudink et al., 2006; Reimert et al.,
2013). Even Morméde et al. (1990) observed that piglets which are kept in a small area
without straw could be more frustrated and stressed in such a housing environment and
therefore display a lower frequency of playing behaviour. Additionally, in the enriched
housing systems, there was more playing observed in the highly stimulating outside areas than
in the inside areas, probably due to the seasonal influences (Newberry et al. 1988), the
unpredictability of the environment (Špinka et al., 2001) and sufficient space for locomotor
play, whereas the pigs might prefer the inside area mainly for resting and sleeping. Moreover,
Martin et al. (2015) claim that enriched housed pigs may develop a broader play repertoire
and the cognitive abilities needed to perform play behaviour earlier than pigs in less enriched
habitats. Consequently, it could be assumed that an enriched habitat leads pigs to augment
their playing behaviour that supposedly reflects the presence of positive affect and good
welfare (Ahloy-Dallaire et al., 2017).
In addition, the feeding method appears to influence playing behaviour. Burghardt (2005)
predicted that enhanced resource availability increases the play levels that correspond to the
lower total durations of playing behaviour (s/h) on farm 1 and farm 2 than on farm 3 in the
final fattening. On farm 1 and farm 2 there was an animal feeding space ratio of 1:4, whereas
farm 3 allocated the largest animal feeding space ratio of 1:1, so it was possible that all pigs
fed simultaneously. Thus, the incidence of playing increased with enhanced resource
52
availability when the pigs weighed over 100 kg (farm 1: 1.20 ± 1.08 (s/h) vs. farm 2: 2.27 ±
1.09 (s/h); farm 3: 2.75 ± 1.08 (s/h)) (Figure 3). Therefore, it might be inferred that housing
systems with adequate feeding space ratios lead to augmented playing behaviour in fattening
pigs, which is again suspected to improve welfare and might be able to generate and by doing
this display positive affective states (e.g. Bateson, 2014; Burgdorf and Panksepp, 2006).
Locomotor play
In the first fattening phase, the pigs on farm 1 (barren habitat) and on farm 3 (enriched
habitat) showed significantly more locomotor play (s/h) than on farm 2 (enriched habitat). In
the final fattening, there is a definite difference between the two housing systems concerning
the occurrence of locomotor play (s/h): the pigs in the barren habitat (farm 1) showed
significantly less locomotor play (s/h) than those in the enriched habitat (farm 2 and farm 3).
As already mentioned, the enriched housing systems offer the three respectively eight times
the space availability for the pigs, especially in the final fattening. This extra-space possibly
extends invitations to the pigs to perform increased locomotor-play that is also reflected by
the longer durations of this playing behaviour (s/h) in the enriched environment (farm 2 and
farm 3) at the final fattening. These findings again emphasise the assumption that playing
behaviour occurs when the environmental conditions are optimal (Lawrence, 1987). In the
first fattening phase, the pigs in the barren environment (farm 1) have approximately double
the space available as they legitimately need (30 - 50 kg: 0.5m² per pig (BMVEL, 2006)) so
they showed total durations of locomotor play (s/h) equivalent to the pigs on farm 3, which
represented one of the enriched habitats. The pigs on farm 2 had the least space availability in
the first fattening phase and therefore presumably performed the least locomotor play at this
level of age. Thus, the opportunity to perform playing, especially locomotor play due to
sufficient space, improves neuromuscular development, motor performance (Byers, 1977) and
cardiovascular fitness (Bekoff 1988; Byers and Walker 1995) resulting in enhanced health
that possibly points to increased welfare. Frequently performed locomotor play improves the
physical condition and authors argue that the better the physical condition in animals, the
more playing results (Sharpe et al., 2002; Nunes et al., 2004; Cameron et al., 2008).
Additionally, these findings might imply that housing conditions, which include sufficient
physical movement opportunities for the pigs, lead to increasing playing behaviour that
simultaneously might provide more possibilities for the developments of positive emotions.
As suggested in previous studies, play can produce affective states in terms of e.g. helping
53
individuals shed negative affective states (Calcagnetti and Schechter, 1992; Held and Špinka,
2011).
Social play
In the first fattening phase it is especially noticeable that the pigs in the barren environment
(farm 1) performed considerably more social play (s/h) than the pigs in the enriched housing
systems (farm 2 and farm 3). In contrast, this definite difference between the two housing
systems was not observed in the final fattening.
In the literature, the occurrence of social playing behaviour has been discussed
controversially. Hence, due to the studies of Hausberger et al. (2012) it could be possible that
the barren housed pigs of the present study showed significantly more playing behaviour
including a large amount of social play in the first fattening phase, because of being
chronically stressed by their low-stimulus environment. Hausberger et al. (2012) examined
the social play behaviour of captive horses and their physiological parameters such as
vertebral disorders, cortisol and oxidative stress values and linked significantly increased
occurrence of social play to poor welfare. In the study, socially playing horses showed more
vertebral disorders, higher oxidative stress values and behaved more aggressively towards
humans than non-social playing horses. Thus, social playing possibly also indicates an
opportunity for animals to cope with their usual, unfavourable life conditions (Hausberger et
al., 2012). This effect might detect a negative affective state of the coping animals, because
animals that are satisfied, e.g. with their environment, do not need to cope, e.g. in an
expression of a behavioural disorder such as belly nosing or sham chewing. As a
consequence, such non-coping animals possibly have a more positive affective state and
experience good welfare (Broom, 1991).
Gender
In the present study, significantly longer total durations of playing behaviour (s/h) occurred in
the female pigs than in the male pigs. This can be explained by different neurological
developments: according to Short and Balaban (1994), females show neurological
development earlier than male pigs and it is said that animals with increased and more
complex dendrite connections in the brain may be more likely to create their own play (Zupan
et al., 2016). Špinka et al. (2001) explained the gender difference based on the many disparate
abilities of motor skills between female and male pigs that influence how they play. It might
54
be also a long-term effect of injury and pain that decreases the playing behaviour of male pigs
due to their castration in early life. For instance, Winder et al. (2017) observed less playing
behaviour in calves in their group home pen after destruction of their horn buds using hot
irons than non-disbudded controls who had undergone a sham procedure. Even these results
may suggest that the lower occurrence of playing behaviour in animals which have recently
experienced negative emotions such as pain possibly indicate affective states including lower
numbers of positive emotions.
Moreover, the present results did not show significant differences in the occurrence of
locomotor play (s/h) between the genders although pursuant to studies of Brown et al. (2015),
there was more locomotor play (s/h) observed in the female pigs than in the male ones.
In this study, the male pigs showed significantly less social play than the females. In contrast,
Brown et al. (2015) observed no significant differences of the occurrence of social play
between the gender even though authors argue that playing behaviour is an important factor in
establishing social relationships in the future (Holmes, 1995), provided that males
traditionally compete for access to females for mating (Graves, 1984).
Age
Bekoff and Byers (1998) suggest that playing behaviour is characteristic of young mammals
and commonly observed at this age of life. Also, Brown et al. (2015) observed that playing
increased in the first six weeks of life and then declined to lower levels by week 14 of life.
The longer total durations of playing behaviour (s/h) on farm 1 and farm 3 in the first
fattening phase than in the final fattening confirm these studies although the difference of
these results on farm 3 did not reach significance. In the first fattening phase, the pigs on farm
1 and farm 3 also showed significantly more locomotor and social play (s/h) than the pigs in
the final fattening. Martin et al. (2015) observed heavier piglets performing less playing
behaviour than lighter piglets after weaning. Maybe the reverse situation on farm 2 is
explained by the holistic change of habitat and a stressful transport by truck as mentioned
above.
Experimental set-up
In addition to the presumed influencing factors on the occurrence of playing behaviour
mentioned, it cannot be excluded that the evaluative subjectivity of each observer also affects
55
the results of the pig’s playing behavioural observation, although intensively repeated courses
of instruction and a cross-classified application for each observer as well as a fixed effect of
the observer in the linear mixed model should minimise this impact. Besides this, the four
observers were not replaced during the whole analysis process of the video data.
CONCLUSION
The aim of this study was to investigate whether the occurrence of playing behaviour in
fattening pigs housed in different housing systems could be useful to obtain a better
understanding of their whole affective state. Following the various approaches of explanation,
playing behaviour might be capable of reflecting quite different emotional states. Thus, it
could be possible that the varied occurrence of playing behaviour in fattening pigs in different
housing systems constitutes a potential indicator of positive as well as negative affective
states including good or poor welfare. Nevertheless, due to the inconclusive results of the
present study and controversially discussed literature, it remains questionable as to whether
the occurrence of playing behaviour is capable of identifying exclusively positive affective
states in fattening pigs. Thus, a combination of further indicators could be useful to
compensate the consisting difficulties in the interpretation.
ACKNOWLEDGEMENTS
The project is supported by funds of the Federal Ministry of Food and Agriculture (BMEL)
based on a decision of the Parliament of the Federal Republic of Germany via the Federal
Office for Agriculture and Food (BLE) under the innovation support programme.
56
REFERENCES
Ahloy-Dallaire, J., Espinosa, J., Mason, G., 2017. Play and optimal welfare: Does play
indicate the presence of positive affective states? Behavioural Processes 156, 3-15.
Alban, L., Petersen, J.P., Busch, M.E., 2015. A comparison between lesions found during
meat inspection of finishing pigs raised under organic/free-range conditions and
conventional, indoor conditions. Porcine Health Management 1:4.
Barnard, C. J., 2004. Mechanism, Development, Function and Evolution. Animal Behaviour,
Upper Saddle River, New Jersey: Pearson Education.
Bateson, P., 2014: Play, playfulness, creativity and innovation. Animal Behaviour Cognition
1, 99-112.
Bateson, P., Young, M., 1981. Separation from the mother and the development of play in
cats. Animal Behaviour, 29, 173-180.
Bekhoff, M. and Byers, J.A., 1998. Animal Play: Evolutionary, Comparative, and Ecological
Perspectives, Cambridge University Press, Cambridge.
Bekoff, M., 1988. Motor training and physical fitness: possible short-comings and long-term
influences on the development of individual differences in behavior. Developmental
Psychobiology, 21, 601-612.
Blokhuis, H., Veissier, I., Jones, B., Miele, M., 2013. Improving farm animal welfare –
Science and society working together: the Welfare Quality approach. In: Blokhuis, H.,
Miele, M., Veissier, I., Jones, B. (Eds.), Improving farm animal welfare. Wageningen
Academic Publishers, Wageningen, Gelderland, Netherlands.
BMVEL, 2006. Neufassung der Tierschutz-Nutztierhaltungsverordnung, Bundesgesetzblatt
Jahrgang 2006 Teil I Nr. 41, ausgegeben zu Bonn am 31.08.2006.
Boissy, A., Manteuffel, G., Bak Jensen, M., Oppermann Moe, R., Spruijt, B., Keeling, L.J.,
Winckler, C., Forkman, B., Dimitrov, I., Langbein, J., 2007. Assessment of positive
emotions in animals to improve their welfare. Physiology and behaviour, 92. 375-397.
Broom, D.M., 1991. Animal welfare: concepts and measurement. Journal of Animal Science,
69, 4167-4175.
57
Brown, S. M.; Klaffenböck, M.; Nevison, I.M.; Lawrence, A. B., 2015. Evidence for litter
differences in play behaviour in pre-weaned pigs. Applied Animal Behaviour Science
172, 17–25.
Brunberg, E., 2011. Tail biting and feather pecking: using genomics and ethology to explore
motivational backgrounds. In: Acta Universitatis agriculturae Sueciae, Dissertation.
Buchenauer, D., 1981. Parameters for measuring welfare, ethological criteria. In: W. Sybesma
(Editor), The Welfare of Pigs. Martinus Nijhoff Publishers, The Hague, 75-89.
Burgdorf, J., Panksepp, J., 2006. The neurobiology of positive emotions. Neuroscience and
Biobehavioral Reviews, 30, 173-187.
Burghardt, G.M., 2005. The Genesis of Animal Play: Testing the Limits. Cambridge,
Massachusetts: MIT Press.
Byers, J.A., 1977. Terrain preferences in the play behavior of Siberian ibex kids (Capra ibex
sibirica). Zeitschrift für Tierpsychologie, 45, 199-209.
Byers, J.A., Walker, C., 1995. Refining the motor play hypothesis for the evolution of play.
American Naturalist, 146, 25-40.
Calcagnetti, D.J., Schechter, M.D., 1992. Place conditioning reveals the rewarding aspect of
social interaction in juvenile rats. Physiology Behaviour 51, 667-672.
Cameron, E. Z., Linklater, W. L., Stafford, K. J., Minot, E. O., 2008. Maternal investment
results in better foal condition through increased play behaviour in horses. Animal
Behaviour 76 (5), 1511–1518.
Chaloupková, H., Illmann, G., BartoŠ, L., Špinka, M., 2007. The effect of pre-weaning
housing on the play and agonistic behaviour of domestic pigs. Applied Animal
Behaviour Science 103, 25-34.
Czycholl, I., Kniese, C., Schrader, L., Krieter, J., 2017. Assessment of the multi-criteria
evaluation system of the Welfare Quality® protocol for growing pigs. Animal. 11,
1573–1580.
D’Eath, R.B., Turner, S.P., 2009. The natural behaviour of the pig. In:Marchant-Fonde, J.N.
(Ed.), The Welfare of Pigs. Animal Welfare, Vol. 7, 13-45.
58
Donaldson, T. M., Newberry, R. C., Špinka, M., Cloutier, S., 2002. Effects of early play
experience on play behaviour of piglets after weaning. Applied Animal Behaviour
Science 79 (3), 221–231.
Dudink, S., Simonse, H., Marks, I., Jonge, F. H. de; Spruijt, B. M., 2006. Announcing the
arrival of enrichment increases play behaviour and reduces weaning-stress-induced
behaviours of piglets directly after weaning. Applied Animal Behaviour Science 101
(1-2), 86–101.
Duncan, I.J.H., 2005. Science-based assessment of animal welfare: farm animals. Revue
Scientifique et Technique – Office International de Epizooties 24(2):483-492.
Fagen, R. M., 1976. Exercise, Play, and Physical Training in animals. In: Bateson P.P.G.,
Klopfer P.H. (eds) Perspectives in Ethology. Springer, Boston, MA.
Fraser, D., Duncan, I. J. H., 1998. ‘Pleasures’,‘pains’ and animal welfare: toward a natural
history of affect. Animal Welfare, 7, 383-396.
Fraser, D., 2008. Understanding animal welfare. Acta Veterinaria Scandinavica 50.
Fraser, D., 2009. Animal behaviour, animal welfare and the scientific study of affect. Applied
Animal Behaviour Science, 118, 108-117.
Gieling, E.T., Nordquist, R.E., van der Staay, F.J., 2011. Assessing learning and memory in
pigs. Animal Cognition. 14, 151-173.
Graves, H.B., 1984. Behavior and ecology of wild and feral swine (Sus crofa). Journal of
Animal Science 58, 482-492.
Hausberger, M., Fureix, C., Bourjade, M., Wessel-Robert, S., Richard-Yris, M.-A., 2012. On
the significance of adult play: what does social play tell us about adult horse welfare?
Naturwissenschaften 99, 291-302.
Held, S.D.E., Špinka, M., 2011. Animal Play and animal welfare. Animal Behaviour 81, 891-
899.
Holmes, W.G., 1995. The ontogeny of littermate preferences in juvenile golden-mantled
ground-squirrels: effects of rearing and relatedness. Animal Behaviour 50, 309-322.
Lawrence, A., 1987. Consumer demand theory and the assessment of animal welfare. Animal
Behaviour 35, S. 293-295.
59
Le Floc'h, N., Matte, J. J., Melchior, D., van Milgen, J., Sève, B., 2010. A moderate
inflammation caused by the deterioration of housing conditions modifies Trp
metabolism but not Trp requirement for growth of post-weaned piglets. Animal: an
international journal of animal bioscience 4 (11), 1891–1898.
Marcet Rius, M., Cozzi, A., Bienboire-Frosini, C., Teruel, E., Chabaud, C., Monneret, P.,
2018. Selection of putative indicators of positive emotions triggered by object and
social play in mini-pigs. Applied Animal Behaviour Science 202, 13–19.
Martin, J. E., Ison, S. H.; Baxter, E. M., 2015. The influence of neonatal environment on
piglet play behaviour and post-weaning social and cognitive development. Applied
Animal Behaviour Science 163, 69–79.
Martin, P., Caro, T.M., 1985. On the functions of play and its role in development. Advances
in the Study of Behavior. Vol. 15 (Ed. By J.L. Rosenblatt & P.J.B. Slater), 59-103.
Orlando, Florida:Academic Press.
Morméde, P., Lemaire, V., Castanon, N., Dulluc, J., Laval, M., Lemoal, M., 1990. Multiple
neuroendocrine responses to chronic social stress – Interaction between individual
characteristics and situational factors. Physiology Behaviour, 47, 1099-1105.
Müller-Schwarze, D., Stagge, B., Müller-Schwarze, C., 1982. Play behavior: persistance,
decrease, and energetic compensation during food shortage in deer fawns. Science,
215, 85-87.
Newberry, R. C.; Wood-Gush, D.G.M.; Hall, J. W., 1988. Playful behaviour of piglets.
Behavioural Processes 17 (3), 205–216.
Nunes, S., Muecke, E.M., Lancaster, L.T., Miller, N.A., Mueller, M.A., Muelhaus, J., Castro,
L., 2004. Functions and consequences of play behaviour in juvenile belding’s ground
squirrels. Animal Behaviour 68, 27-37.
Panksepp, J., 1998. Affective neuroscience: The foundations of human and animal emotion,
New York: Oxford University Press.
Reimert, I., Bolhuis, J.E., Kemp, B., Rodenburg, T.B., 2013. Indicators of positive and
negative emotions and emotional contagion in pigs. Physiology Behaviour, 109, 42-
50.
60
Rushen, J., 2003. Changing concepts of farm animal welfare – Bridging the gap between
applied and basic research. Applied Animal Behavior Science 81:199–214.
SAS 2017: SAS Institute Inc. Cary, NC, USA.
Sharpe, L.L., Clutton-Brock, T.H., Brotherton, P.N.M., Cameron, E.Z., Cherry, M.I., 2002.
Experimental provisioning increases play in free-ranging meerkats. Animal Behaviour,
64, 113-121.
Short, R.V., Balaban, E., 1994. The Differences Between the sexes. Cambridge University
Press, Cambridge, United Kingdom, S. 203-300.
Siviy, S.M., Harrison, K.A., Fear McGregor, L.S., 2006. Risk assessment, and playfulness in
the juvenile rat. Behavioural Neuroscience, 120 (1), S. 49-59.
Siviy, S.M., Panksepp, J., 1985. Energy-balance and play in juvenile rats. Physiology and
Behavior, 35, 435-441.
Špinka, M., Newberry, R. C. & Bekoff, M., 2001. Mammalian play: training for the
unexpected. Quarterly Review of Biology, 76, 141-168.
Temple, D., Dalmau, A., Ruiz de la Torre, J., Manteca, X., Velarde, A., 2011. Application of
the Welfare Quality protocol to assess growing pigs kept under intensive conditions in
Spain. Journal of Veterinary Behavior 6:138–149.
Vanderschuren, L.J.M.J., Niesink, R.J.M., Van Ree, J.M., 1997. The neurobiology of social
play-behavior in rats. Neuroscience and Biobehavioral Reviews, 21, 309-326.
Veissier, I., Aubert, A., Boissy, A., 2012. Animal Welfare – A result of animal background
and perception of its environment. Animal Frontiers 2(3):7–15.
Vinke, C.M., van Leeuwen, J., Spruijt, B.M., 2005. Juvenile farmed mink (Mustela vison)
with additional access to swimming water play more frequently than animals housed
with a cylinder and platform but without swimming water. Animal Welfare 14, 53-60.
Webster J., 2005. The assessment and implementation of animal welfare: theory into practice.
Revue Scientifique Et Technique-Office International Des Epizooties 24:723–734.
Welker, W.I., 1961. An analysis of exploratory and play behavior in animals. In Fiske, D. W.,
and Maddi, S. R. (eds.) Functions of Varied Experience, Dorsey, Homewood, Ill.
61
Winder, C.B., LeBlanc, S.J., Haley, D.B., Lissemore, K.D., Godkin, M.A., Duffield, T.F.,
2017. Clinical trial of local anesthetic protocols for acute pain associated with caustic
paste disbudding in dairy calves. Journal of Dairy Science 100, 6429-6441.
Zupan, M.; Rehn, T.; Oliveira, D. de; Keeling, L. J., 2016. Promoting positive states: the
effect of early human handling on play and exploratory behaviour in pigs. Animal: an
international journal of animal bioscience 10 (1), 135–141.
62
63
CHAPTER THREE
Can tail and ear postures be suitable to capture the affective state of
growing pigs?
K.L. Krugmann , F.J. Warnken, J. Krieter, I. Czycholl
Institute of Animal Breeding and Husbandry, Christian-Albrechts-University Kiel,
Olshausenstr. 40, D-24098 Kiel, Germany.
Submitted to Applied Animal Behaviour Science
64
ABSTRACT
The present study examined whether tail and ear postures in fattening pigs (n=228) housed in
different housing systems could be suitable for assessing their affective state. In doing so, it
investigated the appearance of curled-up, hanging, raised, tucked-under or wagging tails
respectively ears directed forwards, backwards, mixed and laterally. The housing systems
examined included a barren (farm 1, n=138) and two enriched (farm 2, n=54; farm 3, n=36)
habitats that differed concerning the availability of space, enrichment and seasonal influences.
The tails of the pigs were undocked, and the boars were surgically castrated. The pigs’ tail
respectively ear postures were analysed for two days at the beginning and two days at the end
of fattening using the scan sampling method respectively the GLIMMIX procedure to analyse
the data. Especially at the end of fattening, the pigs in the barren housing system showed
significantly fewer curled-up tails than those in the enriched housing system (farm 1: 58.7 ±
11.6 % vs. farm 2: 91.6 ± 2.2 %; farm 3: 95.5 ± 1.8 %). The barren-housed pigs showed also
more raised respectively wagging tails than the enriched-housed pigs (raised tails: farm 1:
25.6 ± 7.6 % vs. farm 2: 1.3 ± 0.6 %; farm 3: 1.4 ± 0.6 %; wagging tails: farm 1: 8.7 ± 4.0 %
vs. farm 2: 3.4 ± 1.1 %; farm 3: 0.2 ± 0.2 %). Particularly at the end of fattening, there were
no definite differences concerning the ears directed forwards between the two housing
systems (end of fattening: Farm 1: 82.3 ± 8.5 %; Farm 2: 63.9 ± 4.9 %; Farm 3: 55.0 ± 6.1 %)
and significantly fewer ears directed laterally were observed in the barren than in the enriched
environment (farm 1: 11.3 ± 4.4 % vs. farm 2: 54.9 ± 3.9 % farm 3: 33.7 ± 3.2 %). Primarily
the curled-up tails could be suitable for indicating a positive affective state of the examined
fattening pigs whereas the other tail postures and principally the ear postures seemed to be
less suitable to reliably represent the pigs’ affective state due to inconclusive results between
the two different housing systems and controversially discussed literature. Nevertheless, the
findings of the present study improve the comprehension of the affective state in fattening
pigs in general.
Keywords: curled-up tails, ear postures, positive emotions, tail postures
65
INTRODUCTION
Consumers are increasingly concerned about the welfare of animals, which contributes to the
decision on what products they buy. Thereby, it is widely accepted that good welfare is not
simply the absence of negative emotions (Boissy et al., 2007). Animal welfare scientists agree
that good welfare includes the promotion and provision of positive affective states and
experiences (Proctor and Carder, 2014). Hence, the study of emotions in animals has been
gaining interest (Panksepp, 1998). There is a wide agreement that animals can feel suffering
and pain and methods for assessing such negative emotions have been developed (Boissy et
al., 2007) but there is ultimately less knowledge of whether animals experience emotions
comparable to humans (Boissy et al., 2007). Nevertheless, scientific investigation of positive
emotions has long been omitted (Boissy et al., 2007). Hence, there is still a lack of established
methods for measuring the animals’ especially positive affective state which includes their
experienced emotions (Fraser, 2009). Therefore, it is an essential issue to become capable of
assessing whether and under what circumstances animals experience particularly positive
emotions (Boissy et al., 2007) and to know exactly which emotions animals can feel and how
they show them (Désiré et al., 2002). However, amongst other things, the subjective character
of short-term affective states as well as long-term mood makes their objective assessment
quite difficult (Curtis and Stricklin, 1991; Rushen, 1996). Hence, the development to measure
affective states as objectively as possible, especially by non-invasive and practicably applied
methods, remains a scientific challenge (Edwards, 2007; Winckler et al., 2007).
LeDoux (1996) states that the best objective and exact way to measure the affective state of
animals is to investigate physiological processes such as hormone transports or
neurotransmitter releases. Further, much more practical ways to assess the affective state of
animals seem to be behavioural observations. Authors suggest that behavioural parameters
such as playing behaviour, affiliative behaviour or vocalisations can be considered as
indicators of positive emotions in animals (Boissy et al., 2007). However, another potential
indicator of the animals’ affective state could be their tail posture, which is known to be
associated with pain or stress in calves and lambs (Graf and Senn, 1999; Grant, 2004). Tail
postures in pigs could conduce as an intraspecific communication method and in canid species
tail postures assist intraspecific signalling in different environments (Kleiman, 1972).
Whereas higher tail postures are suggested to be related to confidence or aggression, a lower
tail posture could reflect fear or submission (Tembrock, 1968; Fox, 1971; Kleiman, 1972;
66
Prince, 1975; Bradbury and Vehrenkamp, 1998). Moreover, curled-up tails in pigs are
assumed to be linked to a positive affective state (Groffen, 2012) and Kiley-Worthington
(1975) concluded curled-up tail postures to be health indicators. Besides, Reimert et al. (2013)
described that tail postures of pigs varied depending on the emotional context. Thus, it might
be conceivable that different tail postures could be useful to gain a better understanding of the
animals’ affective state, including its positive and negative share.
Additionally, Boissy et al. (2011) observed that sheep ears rose up to upright positions during
negative situations whereas positive emotional experiences were related to passive, plane ear
postures. More forward ear postures of sheep were also found in negative experiences such as
social isolation and less in positive situations such as fresh hay feeding (Reefmann et al.,
2009). Additionally, a higher number of ear-posture changes seem to indicate negative
situations in pigs (Reimert et al., 2013). Also cows spend more time keeping their ears in
backward or hanging postures while being stroked, which could point to a more positive
affective state as other authors suggest that cows like to be stroked (Schmied et al., 2010,
2008a, 2008b; Waiblinger et al., 2004; Westerath et al., 2014; Windschnurer et al., 2009).
Hence, as there are highly developed muscles around the ears which ensure they can rotate in
many different ways (Reefmann et al., 2009), Reimert et al. (2013) assumed that ear postures
could be indicative of affective states in pigs.
Thus, the current study aims to promote the general understanding of the body language
signals of fattening pigs especially regarding their tail and ear postures and the associated
context to the pigs’ affective state.
MATERIALS AND METHODS
Animals and housing
The study was conducted on three different farms in Northern Germany over a period from
November 2016 until September 2017. A total of 228 fattening pigs, bred from commercial
cross-bred dams (Large White x Landrace) and sired by Pietrain boars were used. All male
pigs were castrated, and the tails were undocked. The housing systems varied especially
regarding an enriched or barren habitat, the availability of space and the climatic conditions.
67
The pigs in the enriched habitat were weaned at the age of six weeks and those of the barren
housing system were weaned with four weeks of life.
Farm 1 represented a conventional fattening stable (n=138) in a closed system in which the
pigs were housed in groups of 19 in a pen. These pens measured 3.70 x 4.70m resulting in
0.92m² per pig and the floor was half perforated and half concrete. Thus, farm 1 constituted
the barren habitat. There was an animal feeding space ratio of 1:4 and the pigs were fed ad
libitum with pelleted feed through dry-feeding machines. Water was achievable through
nipple drinkers and the ambient temperature was 18°C. Balls, plastic pipes and metal chains
were provided as occupation material and in case of tail-biting, the pigs received gunny sacks.
There was daily artificial light for eight hours (07:00-15:00).
Farm 2 (n=54) depicted an ecological fattening stable with an enriched environment, e.g.
inside and outside pens (in total 4.90 x 9.80m) with straw bedding. The outside pen involved a
roofed (2.70 x 4.90m) and roofless area (3.50 x 4,90m). In the first fattening phase, there were
54 pigs housed in one pen with a total area of 48.02m² (4.90 x 9.80m) occurring in 0.89m² per
pig. After the first fattening phase, the animals were divided into three of these pens evenly so
that each pig had 2.67m² available spaces. The feeding was ad libitum with mealy feed by an
animal feeding space ratio of 1:4.
The third farm (farm 3; n=36) exhibited an ecological fattening stable with an enriched
habitat. There were special soil-based rooting areas, roofed straw-bedded outside pens and
straw-bedded inside pens. The dimension of the whole pen was 83.32m² in total (8.33m² per
pig with a usual occupancy of 10 animals per pen). There was an animal feeding space ratio
of 1:1 and the pigs were fed ad libitum with liquid feeding.
In the enriched housing system (farm 2 and farm 3), there were hay racks and nipple drinkers
available as well but no occupation material as in the barren habitat (farm 1). The ambient
temperature and the daylight length were determined by the season. Supplementary, the pigs
in the enriched housing system got bread and different vegetables and fruits daily.
Experimental design
The present study investigated different tail and ear postures of fattening pigs at the beginning
and at the end of fattening of two batches from three different farms. The different tail and ear
postures were determined by an ethogram (Table 1).
68
Table 1: Ethogram of the different tail and ear postures and presumed, related indicated affective state
Body part Tail/ear
posture
Description Presumed,
indicated
affective state
Reference
Tail
Curled-up
Tail forms a loop above the back of the pig
Positive e.g. McGlone et al. (1990)
Hanging
Tail is neither curled up nor raised but hangs down
Negative/neutral Paoli et al. (2016); Guthrie (1971)
Raised Tail is raised but not curled
Positive/Negative Scheurmann (1974); Reefmann (2009)
Tucked-under Tail is between the hind legs
Negative Groffen (2012)
Wagging Tail is wagging Positive/Negative Kleinbeck and McGlone (1993); Groffen (2012)
Ear
Forwards Ears are directed forwards
Positive/Negative Reefman et al., 2009; Raoult and Gygax (2018)
Backwards Ears are directed backwards
Positive/Negative
Windschnurer et al., 2009; Reimert et al. (2012)
Mixed
One ear is directed forwards and one ear backwards
Negative Reefmann et al. (2009)
Laterally
Ears are directed to the side, neither forwards nor backwards
Positive/Negative Reefmann et al. (2009);
Recording and analysis of the tail and ear postures
All fattening pigs were videotaped two days at the beginning and two days at the end of
fattening using camera systems (HeiTel Digital Video GmbH, Kiel, Germany and AXIS M30-
VE Network Cameras). The video cameras were positioned above the pens to obtain a
complete overview. Each of the pigs was marked individually with a sign on its back. By
using the scan sampling method (from the beginning of the daylight once an hour during the
same) the tail respectively ear posture of each pig was noticed if they were recognisable and
69
the pigs were walking or standing. Tail postures were classified as either “curled-up”,
“hanging”, “raised”, “tucked-under” or “wagging”. Ear postures were determined as “ears
directed forwards”, “ears directed backwards”, “ears mixed” or “ears directed laterally”
(Table 1). The video data were assessed by four observers. These observers were intensively
trained using video test sequences at the beginning of the video analysis and through repeated
courses of instruction between the individual evaluation sections. Additionally, the observers
were applied cross-classified and not replaced during the whole video analysis process.
Statistical analysis
Statistical analyses were performed with the SAS® 9.4 software (SAS Institute Inc., 2017).
The tail respectively ear postures were documented as binomial data (0 = respective tail or ear
posture was not shown; 1 = respective tail or ear posture was shown) during the video
analysis. The data were analysed using the GLIMMIX procedure with a binomial distribution
(link-function = logit). Fixed effects were added to the model in a stepwise manner. The
model for the data of the curled-up tails, hanging tails and ears directed forwards included the
fixed effects farm (1-3), points of testing (beginning respectively end of fattening) nested in
farm, gender (female, male) and a random effect of each individual pig nested in farm. The
model for the data of the raised tails, wagging tails and laterally directed ears included the
fixed effects of farm (1-3), gender (female, male) and a random effect of each individual pig
nested in farm. Statistical significance was determined at p<0.05 and the significance of
differences in the least square means was adjusted with the Bonferroni-correction. Due to rare
occurrences, statistical models could not be used for the data of the tucked-under tails, ears
directed backwards and ears mixed and therefore these results are not shown.
RESULTS
Curled-up tails
Regarding the differences between the farms at each point of testing, there were no significant
differences between the curled-up tails observed on farm 1 and farm 2 at the beginning of
fattening. On farm 1, there were significantly fewer curled-up tails than on farm 3. No
significant differences were observed between farm 2 and farm 3. At the end of fattening,
70
there were significantly fewer curled-up tails on farm 1 than on farm 2 and farm 3 and there
were no significant differences between farm 2 and farm 3 (Table 2).
Concerning the differences between the points of testing within each farm, there were no
significant differences between the curled-up tails at the beginning and the end of fattening on
farm 1. On farm 2 and farm 3, there were significantly fewer curled-up tails at the beginning
than at the end of fattening (Table 2).
Table 2: LS-Means and standard errors of the different tail and ear postures
Tail/ear
posture Point of testing Farm 1 Farm 2 Farm 3
Curled-up
Beginning of fattening
31.5±10.2 % A,a 71.8±7.1 % AB,a 82.4±3.8 % B,a
End of fattening 58.7±11.6 % A,a 91.6±2.2 % B,b 95.5±1.8 % B,b
Hanging
Beginning of fattening
18.7±7.8 % A,a 17.7±5.7 % A,a 14.1±3.1 % A,a
End of fattening 10.4±6.2 % A,a 5.5±1.7 % A,a 3.1±1.5 % A,b
Ears directed
forwards
Beginning of fattening
80.5±7.6 % A,a 3.2±2.3 % B,a 65.7±4.5 % A,a
End of fattening 82.3±8.5 % A,a 63.9±4.9 % A,b 55.0±6.1 % A,a
Raised - 25.6±7.6 % A 1.3±0.6 % B 1.4±0.6 % B
Wagging - 8.7±4.0 % A 3.4±1.1 % AB 0.2±0.2 % B
Ears directed
laterally
- 11.3±4.4 % A 54.9±3.9 % B 33.7±3.2 % C
A, B, C different letters indicate significant differences between farms within each time of testing (p<0.05); a, b different letters indicate significant differences within farms between the different points of testing (p<0.05).
Hanging tails
Relating to the differences between the farms at each point of testing, at both the beginning
and end of fattening, there were no significant differences between the hanging tails observed
on the three different farms (Table 2).
In terms of the differences between the points of testing within each farm, there were no
significant differences between the hanging tails observed at the beginning and the end of
fattening on farm 1 and farm 2. On farm 3, the pigs showed significantly more hanging tails at
the beginning than at the end of fattening (Table 2).
71
Raised tails
Concerning the differences between the farms, there were significantly more raised tails
observed on farm 1, than on farm 2 and farm 3. There were no significant differences between
farm 2 and farm 3 (Table 2).
Wagging tails
In terms of the differences between the farms, there were no significant differences between
the wagging tails observed on farm 1 and farm 2. On farm 1, there were significantly more
wagging tails than on farm 3. No significant differences were observed between farm 2 and
farm 3 (Table 2).
Ears directed forwards
In regard to the differences between the farms at each point of testing, there were significantly
more ears directed forwards observed on farm 1 than on farm 2, at the beginning of fattening.
There were no significant differences between farm 1 and farm 3. On farm 2, there were
significantly fewer ears directed forwards than on farm 3. At the end of fattening, no
significant differences were observed between the farms (Table 2).
Concerning the differences between the points of testing within each farm, there were no
significant differences between the ears directed forwards at the beginning and end of
fattening on farm 1. On farm 2, there were significantly fewer ears directed forwards observed
at the beginning of fattening than at the end. There were no significant differences of the ears
directed forwards between the beginning and end of fattening on farm 3 (Table 2).
Ears directed laterally
Relating to the differences between the farms, there were significantly fewer ears directed
laterally on farm 1 than on farm 2 and farm 3. There were significantly more ears directed
laterally on farm 2 than on farm 3 (Table 2).
Gender
With regard to the differences between the gender, there were significantly more raised tails
observed in the sows than in the castrates (7.1 ± 1.8 % vs. 2.1 ± 1.1 %; p=0.04). A gender
effect occurred exclusively with raised tails.
72
DISCUSSION
The main aim of this study was to investigate whether the occurrence of different tail and ear
postures in fattening pigs in different housing systems is suitable to obtain a better general
understanding of affective state, particularly regarding its positive share. Two distinct
different housing systems were selected, as it was suggested that there might be significant
differences in the appearance of the pigs’ tail and ear postures in these two habitat types, in
order to conclude that the tail respectively ear postures could be able to indicate the pigs’
more positive affective state. It was hypothesised that the fattening pigs with a more positive
affective state were presumably housed on farms 2 and 3 due to a more enriched environment
compared to the barren habitat on farm 1. The enriched environment could be the potential
cause for this difference, as enrichment such as straw bedding or soil-based rooting areas
provide more opportunities for the pigs to perform their natural living, which constitutes an
important part of good welfare (Fraser, 2008) and possibly induces a more positive affective
state.
Curled-up tails
The barren-housed pigs showed significantly fewer curled-up tails than the pigs in the
enriched housing system especially at the end of fattening. Further, the enriched-housed pigs
in studies by Groffen (2012) indicated a curled-up tail more often than barren-housed pigs.
Other authors observed a positive correlation between feeding and drinking with a curled-up
tail. As feeding and drinking is known as positive behaviour that satisfies the essential need
for food and water (Cabanac, 1992; Carver, 2001; Rolls, 2005; Burgdorf and Panksepp, 2006)
it is possible that pigs curl their tails up when they are satisfied, which would indicate that
their affective state is more positive than negative. While there are usually more possibilities
to explore in enriched habitats than in barren housing systems, the curled-up tails observed
more often in the enriched housing system could be, as mentioned above, due to the satisfied
need of the pigs to perform their natural living, which constitutes an important part of animal
welfare (Fraser, 2008). This ability to perform natural behaviour and a simultaneous increase
in the occurrence of curled-up tails in more enriched environments could also indicate that
pigs, when they curl up their tails, are satisfied and therefore presumably possess more
positive affective states. Moreover, Goursot et al. (2018) observed pigs curling up their tails in
rewarding situations, which can be considered as positive experiences and therefore could
also indicate that pigs possess positive affective states when they curl up their tails. Moreover,
73
pigs showed their tails curled up more often in pens without a tail-biting outbreak than in pens
where this behavioural disorder occurred (McGlone et al., 1990). As tail-biting is known to
indicate poor welfare (Edwards, 2011) it is conceivable that the lower appearance of curled-
up tails and the simultaneous occurrence of the described behavioural disorder could point to
a less positive affective state. Also, Zonderland et al. (2009) observed pigs with curled-up
tails often lacking any tail damage, which might also indicate and promote more positive
affective states in these pigs as tail damage is associated with negative emotions such as pain;
presumably it reduces health and therefore might be able to lower the positive share of the
pigs’ affective state. Additionally, Kiley-Worthington (1975) concluded the appearance of
curled-up tails in pigs as a health indicator. Thus, according to the explanations above, it
seems acceptable that the occurrence of curled-up tails clearly points to a more positive
affective state in the enriched- than barren-housed pigs in this study.
Hanging tails
No significant differences between the hanging tails occurred in this study on the farms at the
different points of testing. Except on farm 3 where there were significantly more hanging tails
observed at the beginning than at the end of fattening.
However, according to literature, it could be possible that the occurrence of hanging tails
points to a less positive affective state in pigs. Thus, Paoli et al. (2016) suggest that a hanging
tail presents a defensive response to unwanted tail investigation involving tail-biting. Other
authors have also observed tail-biting occurring mostly when the pigs’ tails hang motionless
(Van Putten, 1980; Arey, 1991). As mentioned above, tail-biting constitutes a behavioural
disorder that points to poor welfare (Edwards, 2011). Moreover, hanging tail postures are also
known to be behavioural patterns of pain in piglets after tail docking (Noonan et al., 1994;
Sutherland et al., 2008). As pain is well known as a negative emotion, Groffen (2012) also
suggested a hanging tail posture to be linked to a more negative affective state. Thus, the
appearance of hanging tails after or during negative experiences could indicate a less positive
affective state in pigs. The significantly more hanging tails on farm 3 at the beginning than at
the end of fattening might be explained by a relative deterioration of their familiar habitat due
to a less attractive stable system compared to their piglet production farms where they were
reared in extensive pasture conditions, possibly resulting in frustrated fattening pigs that did
not feel well and thus performed more hanging tails in the beginning than at the end of
fattening (Le Floc'h et al., 2010).
74
Nevertheless, in ungulates (Guthrie, 1971) and in dogs (Tembrock, 1986; Fox, 1971;
Kleiman, 1972; Prince, 1975; Bradbury and Vehrenkamp, 1998), a hanging tail is seen as a
neutral tail posture, which is assumed to indicate neither positive nor negative affective states.
Hence, it should be noted that it could be difficult to conclude clearly whether hanging tails
indicate exclusively negative or also rather neutral affective states in pigs. However,
presumably, the aforementioned arguments of a hanging tail as an indication of a more
negative affective state might prevail, because a transferability of ungulates and dogs to
fattening pigs with respect to hanging tails could seem to be questionable. Contrary to the
hypothesis in this study, even though no significant differences were observed in the
occurrence of hanging tails between the two housing systems, further studies are needed to
investigate hanging tails as an indication of a more negative affective state in pigs.
Raised tails
There were significantly more raised tail postures observed in the barren housing system than
in the enriched housing system. In the view of Paoli et al. (2016), an upright tail posture
represents a welfare indicator in pigs, which is reduced when tail-biting occurs. Thus, it could
be assumed that a raised tail posture in pigs indicates a more positive affective state.
Moreover, calves (Scheurmann, 1974) and lambs raise their tails while sucking, and other
observations have shown raised tails in sheep while being groomed by humans (Reefmann et
al., 2009). Suspecting that sucking and being groomed are appreciated as positive stimuli, a
raised tail could also be able to indicate a more positive affective state in calves, lambs and
sheep. In contrast to this, raised tails might possibly also indicate a rather negative affective
state in sheep and ungulates. Reefmann et al. (2009) observed sheep raising their tails only
during separation from their herd members and not during rumination or pleasant situations
such as fresh hay feeding. By separation from the group, it is possible that animals feel
negative emotions such as fear and anxiety that promote a more negative affective state.
Additionally, feeding should be a positive experience that satisfies essential needs (Cabanac,
1992; Carver, 2001; Rolls, 2005; Burgdorf and Panksepp, 2006), which confirms the
assumption that more raised tails indicate rather negative affective states in sheep. Moreover,
in ungulates, a raised tail is linked with alarm and flight behaviour (e.g. Alados, 1986), which
could be associated with a more negative affective state as alarm and flight behaviour is
linked to negative emotions such as fear and stress. Additionally, it is said that a raised tail
75
shows a strong emotional activation in general, irrelevant of negative or positive valence
(Reefmann et al., 2009).
However, due to the overall context of this study and a questionable transferability from sheep
and ungulates to pigs, it is difficult to conclude clearly whether a raised tail indicates an
exclusively positive affective state in the examined pigs, although raised tails are assessed as
welfare indicators in pigs in literature. Based on the hypothesis of this study that the pigs with
the more positive affective state were housed on farms 2 and 3 due to their enriched
environment, more raised tails would have had to occur there than in the barren environment.
As this was not the case in this study and also more curled-up tails were observed in the
enriched than the barren environment and in literature curled-up tails clearly seem to be an
indication of a more positive affective state in pigs, the suitability of raised tails as reliable
indicators of a more positive affective state in the examined pigs of this study remains
questionable.
Wagging tails
On farm 1, there were significantly more wagging tails observed than on farm 3. According to
literature, tail wagging increases in food-frustrated situations (Kiley-Worthington, 1975) and
other authors have observed that tail-wagging increases shortly after surgical procedures
(Noonan, 1994; Hay et al., 2003) which are highly stressful and painful (Groffen, 2012).
Moreover, Zonderland et al. (2009) claimed that pigs often show more tail-wagging when
their tails are damaged presumably due to skin irritation (Kiley-Worthington, 1975).
Robertson et al. (1994) assumed that tail-wagging could be an effort to alleviate pain, e.g.
from biting insects or after being bitten by pen mates (Groffen, 2012). In addition to the
described assumptions of tail-wagging as an indication of a more negative affective state in
pigs, it is characterised as a sign of restlessness in cattle (Sylvester et al., 2004) and frustration
in antelope (Kiley-Worthington, 1978). In contrast, Kleinbeck and McGlone (1993) observed
that pigs wag their tail during feeding, which is seen as a positive experience and therefore
could be also associated with a more positive affective state in pigs. Additionally, also in dogs
tail-wagging is known to be linked to a positive affective state (Fatjó et al., 2007).
Furthermore, other authors assume that a wagging tail in black deer simply indicates arousal
and does not have any signalling characteristics (Stuart and Stuart, 1997). Thus, it might be
that also in pigs a wagging tail merely indicates arousal (Groffen, 2012). Hence, due to the
controversial discussions in literature and inconclusive results across the two different
76
housing systems regarding wagging tails in the fattening pigs examined, it is difficult to draw
a clear conclusion as to whether wagging tails could be potential indicators of a pig’s
exclusively more positive affective state.
Ears directed forwards
At the beginning of fattening there were significantly more forward ear postures observed on
farm 1 than on farm 2 and no significant differences occurred between farm 1 and farm 3. On
farm 2, there were significantly fewer ears directed forwards than on farm 3. At the end of
fattening, there were no significant differences between the ears directed forwards on the
three farms.
In studies of Reefmann et al. (2009), more forward ear postures in sheep were found in
negative experiences such as social isolation than in positive situations such as fresh hay
feeding. As social isolation from group members could induce negative emotions such as fear
and anxiety and fresh hay feeding might be able to induce positive affective states because it
satisfies essential needs for feeding and drinking as mentioned above, it could be assumed that
ears directed forwards indicate a less positive affective state in sheep. Also Proctor et al.
(2014) observed cows performing fewer forward ear postures during positive experiences.
However, it is questionable whether ears directed forwards point solely to a negative affective
state as this ear posture could also be linked to attentive animals (Raoult and Gygax, 2018),
which can also be achieved by positive stimuli.
Regarding the entire context of the results of this study and lacking sufficient literature
concerning the emotional context with ears directed forwards in pigs, it seems doubtful
whether the pigs on farm 2 showed conspicuously fewer ears directed forwards due to a much
more positive affective state at the beginning of fattening as described in sheep and cows. In
comparison with the results of the curled-up tails, which most probably indicate a positive
affective state due to clearly discussed literature, it could have been assumed that fewer ears
directed forwards would have also been observed on farm 3. However, since the observations
of the ears directed forwards respectively curled-up tails showed no consistent result and the
latter are clearly discussed as indications of a positive affective state in pigs, it seems as if
ears directed forwards have been hitherto less suitable to represent the affective state in pigs,
whether positive or negative.
77
Ears directed laterally
In the barren housing system, there were significantly fewer ears directed laterally observed
than in the enriched housing system. It seems imaginable that the ears directed laterally in the
present study could be related to both a passive ear posture (ears hanging down relaxed) or a
change in ear postures. Reimert et al. (2013) observed more ear posture changes in pigs in
aversive situations such as social isolation combined with negative, unpredictable
interventions, which could be an indication of a more negative affective state. Additionally, in
studies by Reefmann et al. (2009), fewer ear posture changes in sheep were associated with
positive situations (fresh hay feeding) and more ear posture changes with negative
experiences (social isolation). In opposite to this, higher numbers of passive ear postures in
sheep occurred more in positive than in negative situations (Reefmann et al., 2009).
In the overall view of the present study it might be imaginable that the ears directed laterally
point to a more positive affective state in the enriched-housed pigs, as there were also more
curled-up and fewer wagging tails observed, whereby particularly the former were assessed as
clearly indications of a more positive affective state in pigs. Nevertheless, it should be noted
that due to the snapshot of the scan sampling method, it cannot be unequivocally evaluated as
to whether the ears directed laterally in this study were related to ear posture changes or rather
passive ear postures and both presumably could be able to indicate negative respectively
positive affective states. Furthermore, the pigs’ ear postures could be less suitable to capture
the affective state in general, as there could be breed-specific differences in ear postures.
However, in this study, this can be excluded because the same genetics were housed on each
farm.
Gender
The more raised tails which were observed in the sows than in the castrates might be possibly
explained by external heat signs of the female pigs. Since the examined pigs were slaughtered
with an age of approximately six months, it may be that the sows had already shown first on
heat symptoms such as raised tails.
78
CONCLUSION
The aim of this study was to examine whether different tail and ear postures of fattening pigs
housed in different housing systems could be suitable to obtain a better understanding of their
whole affective state and to identify reliable indicators to assess the pigs’ positive affective
state. Conclusively, primarily the curled-up tail postures might have the potential to assess the
affective state of the examined fattening pigs. Nevertheless, a combined consideration with
further indicators, e.g. physiological ones, would be useful to verify the suitability of the
curled-up tails as reliable indicators. However, the findings of the present study contribute to
the comprehension of the measurement of the positive affective state in pigs in general.
ACKOWLEDGEMENTS
The project is supported by funds of the Federal Ministry of Food and Agriculture (BMEL)
based on a decision of the Parliament of the Federal Republic of Germany via the Federal
Office for Agriculture and Food (BLE) under the innovation support programme.
79
REFERENCES
Alados, C. L., 1986. The use of the tail and rump patch in the dorcas gazelle (Gazella dorcas
L.). Mammalia 50: 439-446.
Arey, D. S., 1991. Tail biting in pigs. Farm Building Progress 105: 20–3.
Boissy, A., Aubert, A., Desire, L., 2011. Cognitive sciences to relate ear postures to emotions
in sheep. Anim. Welf. 20, 47–56.
Boissy, A., Manteuffel, G., Jensen, Bak, M., Randi Oppermann, M., Spruijt, B., Keeling, L.J.,
2007. Assessment of positive emotions in animals to improve their welfare. In:
Physiology & behavior 92 (3), 375–397.
Bradbury, J. W. V., Vehrencamp. S.L., 1998. Principles of animal communication. Sinauer
Associates, Sunderland. MA.
Burgdorf, J., Panksepp, J., 2006. The neurobiology of positive emotions. Neuroscience &
Biobehavioral Reviews. 30, 173–187.
Cabanac, M., 1992. Pleasure:the common currency. Journal of Theoretical Biology 155, 173–
200.
Carver, C. S., 2001. Affect and the functional bases of behavior: on the dimensional structure
of affective experience. Personality and Social Psychology Review, 5, 345–356.
Curtis, S.E., Stricklin, W.R., 1991. The importance of animal cognition in agricultural animal
production systems: an overview. Journal of Animal Science 69, 5001–5007.
Désiré, L., Boissy, A., Veissier, I., 2002. Emotions in farm animals: a new approach to
animal welfare in applied ethology. Behavioural Processes 60, 165–180.
Edwards, S., 2011. What do we know about tail biting today? The pig journal, 81–86.
Edwards, S.A., 2007. Experimental welfare assessment and on-farm application. Animal
Welfare 16, 111–115.
Fatjó, J., Feddersen-Petersen, D., Luís, J.,Torre, R., Amat, M., Mets, M., Braus., B., Manteca,
X., 2007. Ambivalent signals during agonistic interactions. Applied Animal
Behaviour Science 105: 274–283.
80
Fox, M. W., 1971. Behaviour of wolves, dogs, and related canids. Jonathan Cape, London.
Fraser, D., 2008. Understanding animal welfare. Acta Veterinaria Scandinavica 50.
Fraser, D., 2009. Animal behaviour, animal welfare and the scientific study of affect.
Applied Animal Behaviour Science, 118, 108-117.
Goursot, C., Düpjan, S., Tuchscherer, A.; Puppe, B.; Leliveld, L. M. C., 2018. Behavioural
lateralization in domestic pigs (Sus scrofa)-variations between motor functions and
individuals. In: Laterality 23 (5), 576–598.
Graf, B., Senn, M., 1999. Behavioural and physiological responses ofcalves to dehorning by
heat cauterization with or without local anaesthesia. Applied Animal Behaviour
Science 62, 153–171.
Grant, C., 2004. Behavioural responses of lambs to common painful husbandry procedures.
Applied Animal Behaviour Science 87, 255–273.
Groffen, J., 2012. Tail posture and motion as a possible indicator of emotional state in pigs.
Animal Science program, Exam work in Animal Science EX 0567, SLU and
Wageningen University (WUR).
Guthrie, R.D., 1971. A new theory of mammalian rump patch evolution. Behaviour 38,
132—145.
Hay, M., Vulin, A. Genin, S. Sales, P., Prunier. A., 2003. Assessment of pain induced by
castration in piglets: Behavioral and physiological responses over the subsequent 5
days. Applied Animal Behavioral Science 82: 201–218.
Kiley-Worthington, M., 1975. The tail movement of ungulates, canids and felids with
particular reference to their causation and function as display. Behaviour 56: 69–115.
Kiley-Worthington, M., 1978. The causation, evolution and function of the visual displays of
the eland (Taurotragus oryx). Behaviour 66, 179—222.
Kleiman, D. G., 1972. Social behaviour of the maned wolf (chrysocyon brachyurus) and bush
dog (speothos venaticus): A study in contrast. Journal of Mammalogy 53, 791- 806.
Kleinbeck, S., McGlone, J.J., 1993. Pig tail posture: A measure of stress. Technical
University Agricultural Science Technical Report T-5-327: 47–48.
81
Le Floc'h, N., Matte, J. J., Melchior, D., van Milgen, J., Sève, B., 2010. A moderate
inflammation caused by the deterioration of housing conditions modifies Trp
metabolism but not Trp requirement for growth of post-weaned piglets. Animal: an
international journal of animal bioscience 4 (11), 1891–1898.
LeDoux, J., 1996. The emotional brain. The mysterious underpinnings of emotional life, New
York.
McGlone, J.J., Sells, J., Harris, S., Hurst, R.J., 1990. Cannibalism in growing pigs: effects of
tail docking and housing system on behaviour, performance and immune function.
Journal of Agricultural Science and Technology. Report No. T-5-283. Texas Tech
University, 69–71.
Noonan, G. J.; Rand, J. S.; Priest, J.; Ainscow, J.; Blackshaw, J. K., 1994. Behavioural
observations of piglets undergoing tail docking, teeth clipping and ear notching. In:
Applied Animal Behaviour Science 39 (3-4), 203–213.
Panksepp J., 1998. Affective neuroscience. The foundations of human and animal emotions.
New York: Oxford University Press.
Paoli, M. A.; Lahrmann, H. P.; Jensen, T.; D'Eath, R. B., 2016. Behavioural differences
between weaner pigs with intact and docked tails. In: anim welf 25 (2), 287–296.
Prince, J. H., 1975: Languages of the animal world. Thomas Nelson, New York, NY.
Proctor, H. S.; Carder, G., 2014. Can ear postures reliably measure the positive emotional
state of cows? In: Applied Animal Behaviour Science 161, 20–27.
Raoult, C.M.C., Gygax, L., 2018. Valence and Intensity of Video Stimuli of Dogs and
Conspecifics in Sheep: Approach-Avoidance, Operant Response, and Attention.
Animals 8(7), 121.
Reefmann, N., Bütikofer Kaszàs, F., Wechsler, B.; Gygax, L., 2009. Ear and tail postures as
indicators of emotional valence in sheep. In: Applied Animal Behaviour Science 118
(3-4), 199–207.
Reimert, I., Bolhuis, J. E., Kemp, B., Rodenburg, T. B., 2013. Indicators of positive and
negative emotions and emotional contagion in pigs. Physiology and Behavior, 109,
42–50.
82
Reimert, I., Bolhuis, J., Kemp, B., Rodenburg, T., 2012. Indicators of positive and negative
emotions and emotional contagion in pigs. Physiology Behaviour 109, 42–50.
Robertson, I.S., Kent, J.E., Molony, V., 1994. Effect of different methods of castration on
behavior and plasma cortisol in calves of three ages. Research in Veterinary Science.
56.8-17.
Rolls, E. T., 2005. Emotion explained. Oxford, UK: Oxford University Press.
Rushen, J., 1996. Using aversion learning techniques to assess the mental state, suffering,
and welfare of farm animals. Journal of Animal Science 74, 1990–1995.
SAS 2017: SAS Institute Inc. Cary, NC, USA.
Scheurmann, E., 1974. Ursachen und Verhütung des gegenseitigen Besaugens bei Kälbern.
Tierärztliche Praxis. 2, 389–394.
Schmied, C., Boivin, X., Scala, S., Waiblinger, S., 2010. Effect of previous stroking on
reactions to a veterinary procedure. Behaviour and heart rate of dairy cows. Interaction
Studies 11, 467–481.
Schmied, C., Boivin, X., Waiblinger, S., 2008a. Stroking different body regions of dairy cows:
effects on avoidance and approach behavior toward humans. Journal of Dairy
Science 91, 596–605.
Schmied, C., Waiblinger, S., Scharl, T., Leisch, F., Boivin, X., 2008b. Stroking of different
body regions by a human: effects on behaviour and heart rate of dairy cows. Applied
Animal Behaviour Science 109, 25–38.
Stuart, C., Stuart, T., 1997. Field Guide to the Larger Mammals of Africa. Struik, Cape
Town.
Sutherland M.A., Bryer P.J., Krebs, N. and McGlone, J.J., 2008. Tail docking in pigs: acute
physiological and behavioural responses. Animal 2: 292-297.
Sylvester, S. P., Stafford, K. J., Mellor, D. J., Bruce, R. A., Ward, R. N., 2004. Behavioural
responses of calves to amputation dehorning with and without local anaesthesia. Aust.
Vet. J. 82, 697—700.
83
Tembrock, G., 1986. Land mammals. In: T. A. Sebeok (ed.) Animal communication. p 359-
373. Indiana University Press, Bloomington, IN.
Van Putten, G., 1980. Objective observations on the behaviour of fattening pigs. Animal
Regulation Studies 3, 105–18.
Waiblinger, S., Menke, C., Korff, J., Bucher, A., 2004. Previous handling and gentle
interactions affect behaviour and heart rate of dairy cows during a veterinary
procedure. Applied Animal Behaviour Science 85, 31–42.
Westerath, H.S., Gygax, L., Hillmann, E., 2014. Are special feed and being brushed judged
as positive by calves? Applied Animal Behaviour Science 156, 12–21.
Winckler, C., Baumgartner, J., Waiblinger, S., 2007. Perspectives of animal welfare at farm
and group level: introduction and overview. Animal Welfare 16, 105.
Windschnurer, I., Barth, K., Waiblinger, S., 2009. Can stroking during milking decrease
avoidance distances of cows towards humans? Animal Welfare 18, 507–513.
Zonderland, J.J., van Riel, J.W., Bracke, M.B.M., Kemp, B., den Hartog, L.A., Spoolder,
H.A.M., 2009. Tail posture predicts tail damage among weaned piglets, Applied
Animal Behaviour Science, 121, 3-4, 165-170.
84
85
CHAPTER FOUR
Investigation of influence of growing pigs’ positive affective state on
behavioural and physiological parameters using
structural equation modelling
K.L. Krugmann , F.J. Warnken, J. Krieter, I. Czycholl
Institute of Animal Breeding and Husbandry, Christian-Albrechts-University Kiel,
Olshausenstr. 40, D-24098 Kiel, Germany.
Submitted to Journal of Animal Science
86
ABSTRACT
The aim of the present study was to investigate whether the primarily positive affective state
of fattening pigs influences various behavioural and physiological parameters such as the
pigs’ playing behaviour, way of behaving in behavioural tests, body language signals or
diameter and astroglia cell numbers of hippocampi, salivary Immunoglobulin A (IgA) content
or salivary protein composition. Additionally, the suitability of the variables mentioned was
examined to assess the pigs’ positive affective state in practice, which still constitutes a latent
variable not itself measurable. For this, a dataset including behavioural and physiological data
of 60 fattening pigs from three different farms with different housing systems were analysed
by the partial least squares structural equation modelling (PLS-SEM) method. A hierarchical
component model (HCM) was used including the pigs’ positive affective state as a higher-
order component (HOC) and the behavioural and physiological parameters as lower-order
components (LOC). Playing behaviour, body language signals and behavioural tests were
revealed, in this order, to be most influenced by the pigs’ positive affective state since these
resulted in the corresponding path coefficients (PC) of PC=0.83, PC=0.79 and PC=0.62,
respectively. Additionally moderate respectively almost moderate R²-values occurred for the
endogenous latent variables playing behaviour (R²=69.8 %), body language signals (R²=62.7
%) respectively behavioural tests (R²=39.5 %). Furthermore, the indicator of the “locomotor
play” showed the highest indicator reliability (IR) (IR=0.85) to estimate the latent variable of
pigs’ positive affective state. The results of the present study supplement the comprehension
and assessment of the pigs’ positive affective state in general.
Keywords: pig, positive affective state, positive emotions, structural equation modelling
INTRODUCTION
Partial least squares structural equation modelling (PLS-SEM) enables researchers to measure
latent variables, which are not directly assessable while measuring them indirectly using
indicator variables. Thus, it depicts a statistical method that facilitates the estimation and
testing of correlative relationships between dependent and independent variables and the
latent structures in between and aims to explain the highest possible variance of latent
variables by calculating various algorithms (Hair et al., 2017). SEM is used as a common
87
method in social science and marketing concerns, though its popularity has also increased in
agriculture issues (Lamb et al., 2011; Valente et al., 2013) such as milk production (de los
Campos et al., 2006; Heringstad et al., 2009) or animal welfare issues in gallop sports in
horses Müller et al., (2015). Hence, the positive affective state in livestock particularly still
constitutes a latent variable for which there is still no reliable measurement method. However,
there is the hypothesis that the positive affective state could influence different behavioural
and physiological parameters in livestock, so that it might be possible to make statements
about the affective state through the analysis of these possible impacts (e.g. Bosch et al.,
2004; Siviy et al., 2006; Czéh et al., 2006) and whether behavioural or physiological
parameters could be able to assess the positive affective state.
In the first instance, in behavioural tests, animals might react in certain ways in special test
situations, so that their type of behaviour, which e.g. seems to express fear (Forkman et al.,
2007), could possibly lead to conclusions about their affective state. Moreover, it has been
stated that playing behaviour only exists in animals if the animals fare well and feel well
(Siviy et al., 2006) and that it decreases when animals experience negative emotions
accompanied by e.g. threats to their fitness (Fagen 1976; Martin and Caro 1985; Burghardt
2005) or adverse environmental conditions (Müller-Schwartze et al., 1982; Siviy and
Panksepp, 1985) which could also allow conclusions concerning the affective state.
Additionally, the body language signals of animals could also be suitable to indicate their
affective state since e.g. the occurrence of curled-up tails in pigs are assumed to be linked to
their more positive affective state (Groffen, 2012). Moreover, there could be transferable
similarities between humans and pigs, so that it could be reasonable to use physiological
information on human depression research to assess livestock’s and in this case pigs’
primarily positive affective state. For instance, due to results from human depression research
it could be conceivable that pigs with a more positive affective state show higher contents of
salivary Immunoglobulin-A (IgA) (Bosch et al., 2004), larger diameters of hippocampi and
higher numbers of astroglia cells (Czéh et al., 2006) than pigs with a more negative affective
state.
Thus, in the current study, SEM was used to estimate the complex interactions between the
behavioural and physiological latent variables mentioned with regard to the particularly
positive affective state in fattening pigs and its influence on these parameters to obtain
indicators for a better understanding of the affective state of pigs in general.
88
MATERIALS AND METHODS
General scheme of a structural equation model (SEM)
A SEM includes two different types of variables which represent the relationships between
them. There are latent variables, which are not directly measurable, and latent measurement
variables, which make an estimation of the former latent variables possible. A SEM is formed
by a structural model (endogenous respectively exogenous latent variables) and at least two
measurement models (latent measurement variables and their self-characterising indicators).
The relationships (path coefficients (PC)) between the endogenous respectively exogenous
latent variables indicate the structural model by arrows connected to each other. Endogenous
(dependent) latent variables can be influenced by other latent variables, irrelevant of being
endogenous or exogenous, although exogenous (independent) latent variables cannot be
affected by other latent variables. Thus, the coefficient of determination (R²), which indicates
to which extent the latent variable is explained by other latent variables, is exclusively
computed for endogenous latent variables (Figure 1).
Hypotheses of the present SEM
This study examined the hypotheses that the primarily positive affective state of fattening
pigs’ influences behavioural and physiological parameters and that behavioural and
physiological parameters could be useful to estimate the pigs’ positive affective state.
Moreover, the positive affective state of fattening pigs depicts the latent variable, which is not
directly measurable, and the behavioural and physiological parameters form the latent
measurement variables, which enables an estimation of the pigs’ positive affective state. Thus,
the positive affective state is an exogenous latent variable and the behavioural and
physiological parameters are endogenous latent variables, which are influenced by the pigs’
positive affective state. According to literature, the animals’ affective state consists of
different emotions such as pleasure, happiness, pain and suffering and other feelings such as
hunger and thirst that are experienced as pleasant or unpleasant (Fraser, 2008). Moreover, the
primarily positive affective state presumably includes experienced pleasant emotions such as
happiness (Ortony and Turner, 1990; Diener and Lucas, 2000) whereas unpleasant emotions
such as fear, pain or suffering probably indicate the animals’ more negative affective state.
(a)
(b)
Figure 1: (a) general scheme of a structural equation model (SEM). IR=indicator reliability; (b) reflective-reflective hierarchical component model (HCM). IR=indicator reliability.
90
Hierarchical Component Models (HCM)
The section above “General scheme of a structural equation model (SEM)” described first-
order models, which contain a single level of constructs. However, in other situations, as in
this study, very complex constructs have to be analysed. Such constructs require an
operationalisation at a higher abstraction level, so a hierarchical component model (HCM)
with two levels of abstraction was developed to analyse the construct of this study. The two
levels of abstraction of an HCM are the higher-order component (HOC), which assesses the
more abstract level, and the lower-order components (LOC), which include the sub-
dimensions of the higher component. Further, the pig’s affective state depicts the HOC, and
the different behavioural and physiological parameters represent the LOC. A reflective-
reflective HCM which contained a reflective relationship between the HOC and LOC was
used whereby all constructs of LOC were specified by reflective measurement models. A
reflective measurement model means that the indicators represent the underlying construct
and that the causality of the construct is directed to the indicators. In order to form the HOC
measurement model, all indicators of the LOC are also assigned to the HOC (Hair et al.,
2017) (Figure 1).
Assessment of the SEM
The SmartPLS 3.0 software (Ringle et al., 2015) was used to generate and calculate the SEM.
The evaluation was based on a variety of recommended quality criteria by Hair et al. (2017).
This evaluation involves a two-stage process whereby firstly the measurement models are
checked, followed by the structural model. Quality criteria of the measurement models
included the internal consistency reliability (composite reliability (CR)) and convergent
validity (indicator reliability (IR) and average variance extracted (AVE)). For the consistency
reliability, the composite reliability considers different loadings of indicators and displays
values between 0 and 1 whereas it should assume values between 0.6 and 0.7 in exploratory
research studies. For the convergent validity, the indicator reliability assesses how sufficient a
latent variable is estimated by an indicator. Thus, indicators with indicator reliabilities
between 0.4 and 0.7 should only be removed from the construct during the assessment process
if their removal increases the quality criteria of the model (Hair et al., 2017), whereas
indicators with indicator reliabilities below 0.4 should always be removed from the model
(Bagozzi et al., 1991; Hair et al., 2011). Further, the elimination of indicators should be
carefully considered as they should be retained if they make a corresponding contribution to
91
the content validity of the whole model (Hair et al., 2017). AVE describes the extent to which
a latent construct explains the variance of its indicators and AVE should achieve values above
or equal to 0.5.
Primary quality criteria of the structural model involve the relevance and statistical
significance of path coefficients and the examination of the coefficients of determination (R²).
The path coefficients indicate the relationships between the constructs in the structural model
and correspond to the standardised regression coefficients in the regression analysis. Path
coefficients show values between -1 and 1 and verify or refute the pre-hypothesised
relationships between the constructs in the model. For this, sign and magnitude have to be
considered (Henseler et al., 2009). An estimated path coefficient near to +1 represents a
strong positive relationship between two examined constructs (and vice versa for negative
values), which is typically statistically significant. The closer the estimated path coefficient is
to 0, the weaker the relationship. Very low values close to 0 are generally not statistically
significant (Hair et al., 2017) whereby the statistical significance is tested by the
bootstrapping method. The coefficients of determination (R²) indicate the proportion of the
variance of an endogenous construct that is explained by all precursor constructs associated
with the endogenous construct. The higher the R² is, the better the construct is explained by
the latent variables in the structural model, which show structural path model relationships on
the explained construct (Hair et al., 2017). In scientific marketing research, R2-values of 0.25,
0.50 and 0.75 are assessed as weak, moderate and substantial, respectively (Henseler et al.,
2009; Hair et al., 2011).
Split-half scenario
To further assess the validity of the SEM, the split-half consistency was applied (Martin and
Bateson, 2007). After the SEM had been applied to the original dataset, this original dataset
was split into two halves by using the PROC SURVEYSELECT procedure in SAS® 9.4 (SAS
Institute Inc., 2017), to which the SEM was also applied.
Data origin and data structure
The dataset of the present study originated from 60 fattening pigs of two batches from three
different farms with two different housing systems in Northern Germany. SAS® 9.4 (SAS
Institute Inc., 2017) was used for the calculation of descriptive statistic. The dataset (n=60)
included behavioural data of two behavioural tests (human approach test (HAT) and novel
92
object test (NOT)), playing behaviour (total duration, locomotor play and social play), body
language signals (curled-up, hanging, raised and wagging tails as well as ears directed
forwards, backwards, mixed and laterally) and physiological data of the saliva regarding its
Immunoglobulin A (IgA) content and protein composition (general protein content (GPC),
total number of bands (TNB) and total band intensity (TBI)) as well as physiological data of
the diameter and astroglia cell number of the hippocampi. The behavioural tests included
parameters of the approach latency (AL), duration of contact (DC) and number of contacts
(NC) in each test. AL represented the time in seconds that each pig needed to approach the
unknown human (in the HAT) or the novel object (in the NOT) until the snout touched the
human or the novel object. DC exposed the accumulated seconds in which the pigs touched
the human or the novel object with their snouts. The entire number of snout contacts that
occurred during the test phase was indicated by NC. The whole test phase was three minutes
and each pig was tested alone in the home pen. The unknown human in the HAT was always
a female person who the pigs did not know from daily routine work. A plastic duck presented
to the pigs in three different sizes (according to the age and the live weight of the animals)
represented the utilized novel objects in the NOT. These ducks showed a yellow body colour
with a red coloured beak. The behavioural data of the behavioural tests were recorded at the
beginning, middle and at the end of fattening and then summarised for each individual pig.
The playing behaviour involved parameters such as the total duration of playing behaviour,
locomotor play and social play in seconds per hour during the daylight of each pig. The body
language signals represented the prevalence of the curled-up, hanging, raised and wagging
tails as well as ears directed forwards, backwards, mixed and laterally once an hour during the
daylight of each individual pig. The behavioural data of playing behaviour and body language
signals were recorded for two days at the beginning and two days at the end of fattening. The
hippocampi diameters were measured in millimetres and the astroglia cell number in GFAP
pixel intensity. The IgA content (µg/ml) of the pigs’ saliva were analysed with a direct
quantitative sandwich-ELISA-Kit for pig-IgA (Celltrend GmbH, Luckenwalde, Germany).
The GPC (µg/ml) were analysed by colorimetric detection by means of a bicinchoninic acid-
based protein assay kit (Pierce™ BCA Protein Assay Kit, ThermoFisher Scientific Inc.,
Waltham, USA) and the TNB and TBI (px) through one-dimensional SDS-polyacrylamide gel
electrophoresis. All physiological data were collected merely at the end of fattening (Table 1).
During the selection process of data, it was considered that each variable contained less than
15 % missing values (Hair et al., 2014).
93
The other two datasets (n=30; n=30) which were used for the split-half scenario included 30
randomly selected animals from the original dataset, respectively (PROC SURVEYSELECT).
Table 1: Descriptive data of indicators used in the SEM
Related
latent
variable
Indicator n
Mean Median Minimum Maximum Standard
deviation
Behavioural
tests
(endogen)
HAT-AL (s) 60 250.90 217.00 2.00 540.00 153.56 HAT- DC (s) 60 138.32 103.50 0.00 419.00 127.26 HAT-NC (s) 60 4.25 4.00 0.00 13.00 2.55 NOT-AL (s) 60 85.07 38.50 5.00 386.00 97.36 NOT-DC (s) 60 179.28 160.00 21.00 459.00 104.24 NOT-NC (s) 60 9.38 9.00 2.00 16.00 2.88
Playing
behaviour
(endogen)
Playing behaviour
(s/h) 60 27.81 20.67 0.00 112.88 26.50
Locomotor play
(s/h) 60 8.89 7.50 0.00 42.05 9.37
Social play (s/h) 60 18.88 10.32 0.00 102.88 22.26
Body
language
signals
(endogen)
Curled-up tails
(prevalence) 60 0.03 0.03 0.00 0.13 0.03
Hanging tails
(prevalence) 60 0.01 0.00 0.00 0.06 0.02
Raised tails
(prevalence) 60 0.01 0.00 0.00 0.13 0.02
Wagging tails
(prevalence) 60 0.00 0.00 0.00 0.06 0.01
Ears directed
forwards
(prevalence)
60 0.03 0.02 0.00 0.19 0.03
Ears directed
backwards
(prevalence)
60 0.00 0.00 0.00 0.01 0.00
Ears directed mixed
(prevalence) 60 0.00 0.00 0.00 0.02 0.00
Ears directed
laterally
(prevalence)
60 0.02 0.01 0.00 0.09 0.02
Slaughter
organs
(endogen)
Diameter of
hippocampi (mm) 60 5.27 5.38 4.25 6.50 0.52
Astroglia cell
numbers (GFAP
pixel intensity)
60 175.82 174.50 0.02 0.07 0.01
Saliva
(endogen)
IgA (µg/ml) 59 125.46 71.33 8.61 992.30 154.19 GPC (µg/ml) 60 1513.58 1185.64 341.97 4563.21 923.17 TNB 60 11.90 12.00 5.00 17.00 2.20 TBI (px) 60 392.86 381.65 146.31 1053.33 190.56
94
RESULTS
The subsequent sections present the results of the final SEM. An initial SEM was calculated
and then checked regarding quality criteria of measurement respectively structural models
which are explained in the previous section “Materials and methods”. To prove validity of this
final SEM, it was also applied to a second dataset (split-half scenario).
Initial SEM
The initial SEM estimated five relationships between six latent variables, divided into five
endogenous latent variables (LOC) and one exogenous latent variable (HOC). Thus, five
arrows connected one exogenous latent variable with five endogenous latent variables, which
served as latent measurement variables associated with 23 indicators (Figure 2).
Final SEM
In the final SEM, three relationships between four latent variables remained (three
endogenous latent variables (LOC) and one exogenous latent variable (HOC)). According to
the quality criteria of the indicator reliability, 17 of the initial 23 indicators were removed
from the model during the assessment process, as their loadings, which represent the
magnitude of indicator reliability, showed values below 0.4 or did not achieve statistically
significance (Hair et al., 2017). The endogenous measurement latent variable slaughter
organs, respectively saliva and their remaining associated indicators “astroglia cell numbers”
respectively “total number of bands” were in case of becoming a so-called single-item also
removed from the SEM as the use of single-items is risky, because it generally reduces
reliability and negatively impacts prediction validity (Hair et al., 2017) (Figure 3).
Assessment of the measurement models (final SEM)
The SEM measurement models were checked for reliability and validity based on the test
criteria composite reliability, indicator reliability and average variance extracted, which are
explained in “Materials and methods”. After calculating the first PLS-algorithm, all indicators
with loadings below 0.4 or without statistically significance were removed from the initial
model (Hair et al., 2017). For behavioural tests, the associated indicator “NOT-approach
latency” showed higher indicator reliability (IR=0.80) than the indicator “HAT-approach
latency” (IR=0.78). Thus, “NOT-approach latency”, appeared to be better than “HAT-
approach latency” to estimate behavioural tests. For playing behaviour, the indicator
95
“locomotor play” (IR=0.93) estimated the latent variable more sufficiently than the indicator
“total play” (IR=0.84). The relating indicators “curled-up tails” and “ears directed laterally”
of the body language signals indicated similar indicator reliabilities (IR=0.90 respectively
IR=0.90), which means that these two estimate body language signals equally well. For the
pig’s positive affective state, “locomotor play” (IR=0.85) was the most appropriate indicator
to estimate its latent variable, followed by “curled-up tails” (IR=0.71) and “ears directed
laterally” (IR=0.71), “total play” (IR=0.59), “NOT-approach latency” (IR=0.51), and “HAT-
approach latency” (IR=0.48) (Table 2).
As a next step, the reliability of the measurement models was examined using the composite
reliability. The measurement models behavioural tests had the lowest composite reliability
(CR=0.77) whereas playing behaviour and body language signals had the highest composite
reliabilities (CR=0.88 respectively CR=0.89). The composite reliability of the pigs’ positive
affective state (CR=0.81) was in between the above-mentioned values. As the composite
reliability was above 0.6 for all measurement models, all measurement models are proven to
be reliable (Table 2).
The test criterion average variance extracted was used to examine validity of the measurement
models. For all measurement models apart from pigs’ positive affective state, the average
variance extracted was above 0.5 (behavioural tests: AVE=0.63; playing behaviour:
AVE=0.79; body language signals: AVE=0.81); this proves the validity of all these
measurement models (Henseler et al., 2009). The value for the pig’s affective state
(AVE=0.43) is the lowest of all the measurement models and was close to the set threshold
(Table 2).
Assessment of the structural model (final SEM)
The evaluation of the structural model revealed that all relationships between the exogenous
and endogenous latent variables were statistically significant (p<0.05). Playing behaviour was
the most influenced parameter by the pigs’ positive affective state by indicating a path
coefficient of 0.83. After that, body language signals were more influenced by the pigs’
positive affective state than behavioural tests (PC=0.79 respectively PC=0.62).
The R²-values of the structural model were assessed in the following step. Here, the structural
model explained 39.5 % of the variance in behavioural tests, 62.7 % of the variance in body
language signals and 69.8 % of the variance in playing behaviour. Therefore, the structural
96
model best explains the variance of playing behaviour, followed by the variance of body
language signals and the variance of behavioural tests (Table 2).
Assessment of the split-half scenario SEM
Also the split-half scenario SEM1 and split-half scenario SEM2 were checked for reliability
and validity. As shown in Table 2, the test criteria of the split-half scenario SEMs indicated
similar tendencies as the results of the final SEM. The relationships (path coefficients)
between the exogenous and endogenous latent variables were statistically significant as well
(p<0.05).
Table 2: Indicator reliabilities (IR), coefficients of determination (R²), composite reliabilities (CR) and average variance extracted (AVE) of the final SEM, split-half consistency SEM1 and split-half consistency SEM2
Final SEM Split-half consistency SEM1 Split-half consistency SEM2
Latent
variable
Indicators IR CR AVE R² IR CR AVE R² IR CR AVE R²
Behavioural
tests
(endogenous)
HAT-
approach
latency
0.78
0.77 0.63 0.39
0.82
0.71 0.55 0.39
0.87
0.82 0.69 0.53 NOT-
approach
latency
0.80 0.65 0.80
Playing
behaviour
(endogenous)
Total play 0.84 0.88 0.79 0.69
0.91 0.92 0.85 0.77
0.76 0.84 0.72 0.69 Locomotor
play 0.93 0.93 0.93
Body
language
signals
(endogenous)
Curled-up
tails 0.90
0.89 0.81 0.62 0.85
0.86 0.76 0.60 0.88
0.87 0.78 0.61 Ears directed
lateral 0.90 0.89 0.88
Pigs’ positive
affective
state
(exogenous)
HAT-
approach
latency
0.48
0.81 0.43 -
0.52
0.82 0.44 -
0.66
0.82 0.44 -
NOT-
approach
latency
0.51 0.39 0.54
Total play 0.59 0.77 0.49 Locomotor
play 0.85 0.85 0.86
Curled-up
tails 0.71 0.62 0.68
Ears directed
laterally 0.71 0.73 0.69
Figure 2: Initial SEM. Circles indicate latent variables; rectangles represent indicators; arrows connecting circles represent relationships between latent variables (path coefficients); arrows between circles and rectangles indicate indicator reliabilities
Figure 3: Final SEM. Circles indicate latent variables; rectangles represent indicators; arrows connecting circles represent relationships between latent variables (path coefficients); arrows between circles and rectangles indicate indicator reliabilities; numbers in circles represent the coefficient of determination.
100
DISCUSSION
Assessment of the measurement models (final SEM)
Concerning indicator reliability, six of the initial 23 indicators were considered reliable for the
estimation of the respective associated latent variables. The high loadings (above 0.7) of the
indicators of the endogenous latent variables indicate that they all seem to be well suited to
measure their latent variable. The indicator reliabilities of the exogenous latent variable pigs’
positive affective state indicated values below 0.7 for the indicators “total play”, “NOT-
approach latency”, and “HAT-approach latency”. Nonetheless, these indicators were not
removed from the model, as indicators with indicator reliabilities between 0.4 and 0.7 should
only be removed from the construct if their removal increases the quality criteria of the model
(Hair et al., 2017). Further, the elimination of indicators should be carefully considered as
they should be retained if they make a corresponding contribution to the content validity of
the whole model (Hair et al., 2017), which was the case in this study. However, according to
the highest indicator reliability, the indicator “locomotor play” seems to be most appropriate
to measure the latent variable of the pigs’ positive affective state, followed by the indicators
“curled-up tails” and “ears directed laterally”, “total play”, “NOT-approach latency” and
“HAT-approach latency”.
These results seem to be realistic, as the occurrence of playing behaviour revealed to be
suitable to show a more positive affective state of the examined pigs. Also found in other
studies, playing behaviour only exists if animals fare well and feel well (Siviy et al., 2006)
and decreases when animals experience negative emotions accompanied by e.g. threats to
their fitness (Fagen 1976; Martin and Caro 1985; Burghardt 2005) or adverse environmental
conditions (Müller-Schwartze et al., 1982; Siviy and Panksepp, 1985). The fact that the
indicator “locomotor play”, according to the indicator reliabilities, seems to be the most
suitable for estimating the pigs’ positive affective state, could be related to the fact that the
opportunity to perform locomotor play improves neuromuscular development, motor
performance (Byers, 1977) and cardiovascular fitness (Bekoff 1988; Byers and Walker 1995)
resulting in enhanced health. This possibly points to increased welfare including experienced
positive emotions as basic health and biologic functioning depicts one of the most important
parts of animal welfare (Fraser, 2008) and affirms the presumably good suitability of
“locomotor play” to estimate the pigs’ positive affective state. However, the indicator “total
play”, which consists of the locomotor play and social play of the examined pigs, indicated
101
lower indicator reliability than “locomotor play”. This could be explained by the
controversially discussed literature regarding the occurrence of social play as an indication of
an exclusively positive affective state. Hausberger et al. (2012) linked the significantly
increased occurrence of social play in captive horses to poor welfare and therefore to a more
negative affective state. Thus, there is no definite evidence that social play indicates an
exclusively more positive affective state in the fattening pigs examined, even though the
transferability from horses to pigs could be questionable. However, it might explain the lower
indicator reliability of “total play” compared to “locomotor play”. Nevertheless, both
indicators of playing behaviour can be concluded as suitable indicators to estimate the latent
variable of the pigs’ positive affective state.
Hence, particularly the indicator “curled-up tails” of the body language signals seems to be
realistic to measure the pigs’ positive affective state, as according to literature pigs curl up
their tails in positive situations such as satisfying their essential needs for feeding and
drinking (Cabanac, 1992; Carver, 2001; Rolls, 2005; Burgdorf and Panksepp, 2006),
rewarding events (Goursot et al., 2018) or being housed in enriched environments (Groffen,
2012). Kiley-Worthington (1975) interpreted the appearance of curled-up tails in pigs even as
health indicators. Thus, it seems acceptable that curled-up tails could be suitable to estimate
the pigs’ positive affective state since satisfying essential needs, rewarding events, enriched
environments or good health presumably induce positive emotions and therefore a positive
affective state in pigs. As mentioned above, good health and biologic functioning represent
one of the most important part of animal welfare. The given opportunity to perform natural
living in enriched environments probably indicate enhanced welfare. In consequence, the
pigs’ positive affective state as natural living constitutes an important part of good welfare as
well (Fraser, 2008). According to the calculated indicator reliabilities, the ears directed
laterally also seem to be suitable to estimate the pigs’ positive affective state almost as well as
the curled-up tails. However, regarding this, further research is needed since it is imaginable
that the ears directed laterally of the present study could be related to both a passive ear
posture (ears hanging down, relaxed) or a change in ear postures. Further, Reimert et al.
(2013) observed more ear posture changes in pigs in aversive situations such as social
isolation combined with negative, unpredictable interventions, which could be an indication
of a more negative affective state. In opposite to this, higher amounts of passive ear postures
in sheep occurred more in positive than in negative situations (Reefmann et al., 2009). Thus, it
is difficult to conclude whether the ears directed laterally in the present study estimate an
102
exclusively positive affective state since due to the snapshot of the scan-sampling method, it
cannot be unequivocally evaluated whether the ears directed laterally in this study are related
to ear posture changes or rather passive ear postures. Both, presumably, could be able to
indicate negative respectively positive affective states.
Moreover, the indicators “HAT-approach latency” and “NOT-approach latency” of the
behavioural tests could be realistic to assess the pigs’ affective state. According to Brown et
al. (2009), quicker latencies to approach novel stimuli, such as unknown humans or novel
objects, are associated with less fearful animals. Therefore, it could be imaginable that
quicker approach latencies indicate pigs which possess a more positive affective state due to
less fearfulness. However, quicker approach latencies could also be a reason for a stronger
motivation to explore novel stimuli (Stolba and Wood-Gush, 1980) related to boredom in the
housing environment. This probably indicates pigs with a more negative affective state, since
then the pigs are possibly not able to perform their natural living, which depicts an important
part of good welfare (Fraser, 2008), as mentioned above. Based on these results and also due
to the controversially discussed literature concerning the ears directed laterally as well as
social play, it could also be assumed that the exogenous latent variable pigs’ positive affective
state corresponds more to the general affective state than to the primarily positive one.
Additionally, it seems quite understandable that according to the higher indicator reliability,
the indicator “NOT-approach latency” could be more appropriate to estimate the pigs’
positive affective state than the indicator “HAT-approach latency” as the pigs’ behaviour in
human approach tests can be influenced by their previous experiences with humans. There are
studies which demonstrate that negative handling leads to more avoidance responses (Carreras
et al., 2017). Hence, it could be conceivable that the pigs’ behaviour in the NOT is better
suited than the pigs’ behaviour in the HAT to estimate the pigs’ positive affective state.
Presumably, in further studies, it would be useful to include further indicators of behavioural
tests in the dataset, for instance related observations between quick approach latencies or
avoidance reactions, which could provide more detailed information on whether the pigs show
quick approach latencies due to less fear or a high motivation to explore unknown humans or
novel objects.
All values were above 0.6 for composite reliability, which indicates reliability in general
(Henseler et al., 2009). Thus, all four latent variables were sufficiently measured by their
respective indicators.
103
Three out of the four measurement models reached the recommended threshold for average
variance extracted, which states that AVE should be above than 0.5 (Henseler et al., 2009).
The average variance extracted for the exogenous latent variable pigs’ positive affective state
was 0.43, which is close to the mentioned threshold. An explanation for the lower average
variance extracted could be the lower indicator reliabilities of the indicators “HAT-approach
latency” (IR=0.48), “NOT-approach latency” (IR=0.51) and “total play” (IR=0.59).
Therefore, it could be possible that the pigs’ positive affective state is not sufficiently
estimated by these indicators, which could also indicate an insufficient estimation of the
whole exogenous latent variable of the pigs’ positive affective state. This, however, seems
obvious since this study is a first approximation, which does not exist in the form yet and
which lays the foundation for making the pigs’ affective state almost measurable and
understandable in the future. Conclusively, as mentioned above, additional indicators are
necessary to complement the estimation of the pigs’ positive affective state sufficiently. Also
Ringle (2004) advises modifying the measurement model when there are deficiencies in
discriminant validity, indicated by average variance extracted values below 0.5.
Assessment of the structural model (final SEM)
All three relationships between the exogenous and the endogenous latent variables showed
statistical significance according to the bootstrapping method. Thus, the pigs’ positive
affective state influenced all three endogenous latent variables significantly. Moreover, the
highest path coefficient between the pigs’ positive affective state and playing behaviour
indicated that this was the most influenced parameter (PC=0.83), followed by the body
language signals (PC=0.79) and behavioural tests (PC=0.62). This is in line with expectations
since also in previous literature playing behaviour has been suggested as suitable to indicate
good welfare and a positive affective state (Fagen, 1976; Lawrence, 1987; Boissy et al.,
2007). The results of path coefficients also show that the body language signals such as
curled-up tails respectively ears directed laterally as well as the pigs’ type of behaviour in the
behavioural tests are influenced by the pigs’ (positive) affective state. Hence, these results
also suggest that playing behaviour, the body language signals mentioned, and behavioural
tests could be potential indicators of a realistic assessment of the pigs’ positive affective state
in practice. All other parameters that were eliminated in the quality assessment of the SEM
therefore do not appear to be significantly influenced by the pigs’ positive affective state, and
therefore do not appear to be suitable for capturing the pigs’ positive affective state in
104
practice. This seems to be comprehensible especially with regard to the elimination of the
salivary parameters during the quality assessment of the model, since it is conceivable that the
saliva of the pigs may be affected by many factors such as different ingredients in the feed,
and especially the salivary IgA-content may rather be influenced by the health status of pigs
than by their positive affective state.
Concerning the coefficients of determination (R²), the model explained moderate R²-values
for the endogenous latent variables playing behaviour (69.8 %) and body language signals
(62.7 %) and an almost moderate R²-value for the endogenous latent variable behavioural
tests (39.5 %) (Henseler et al., 2009; Hair et al., 2011). This proves that the pigs’ positive
affective state explains the variance of playing behaviour and body language signals slightly
better than the variance of the pigs’ behaviour in the behavioural tests and could imply that
the former could be preferable to assess the pigs’ positive affective state in practice.
Additionally, it could be conceivable that behavioural tests are also influenced by the animals’
personality traits (Boivin et al., 1992) and not only by the affective state. However, further
indicators should be tested in the future and researched in other animal species to obtain a
more detailed understanding of the affective state.
Comparison of the final SEM with the split-half scenario SEM
Since the results of the split-half scenario SEMs indicated similar tendencies of test criteria
for reliability and validity as the final SEM does (Table 2), the model quality was assessed as
preserved. This confirms the overall validity of the final SEM that was used to pursue the
research issue of this study.
CONCLUDING STATEMENT
The current study is one of the first research studies using the PLS-SEM method for the
assessment of the pigs’ particularly positive affective state and its possible influence on
various behavioural and physiological parameters. This method seems to be suitable to
analyse the influence of the pigs’ positive affective state on behavioural and physiological
parameters and to assess the pigs’ positive affective state in practice. However, this study
which contributes to the understanding of the affective state in pigs in general depicts a first
approximation to estimate the pigs’ positive affective state using the PLS-SEM method.
105
ACKNOWLEDGEMENTS
The project is supported by funds of the Federal Ministry of Food and Agriculture (BMEL)
based on a decision of the Parliament of the Federal Republic of Germany via the Federal
Office for Agriculture and Food (BLE) under the innovation support programme.
106
REFERENCES
Bagozzi, R.P., Yi, Y., and L.W. Philipps, 1991. Assessing construct validity in organizational
research, Administrative Science Quarterly, 36, 421-458.
Bekoff, M., 1988. Motor training and physical fitness: possible short-comings and long-term
influences on the development of individual differences in behavior. Developmental
Psychobiology, 21, 601-612.
Boissy, A., Manteuffel, G., Bak Jensen, M., Oppermann Moe, R., Spruijt, B., Keeling, L.J.,
Winckler, C., Forkman, B., Dimitrov, I., Langbein, J., 2007. Assessment of positive
emotions in animals to improve their welfare. Physiology and behaviour, 92. 375-397.
Boivin, X., Le Neindre, P., Chupin, J.M., 1992. Establishment of cattle–human relationships.
Applied Animal Behaviour Science 32, P. 325–335.
Bosch, J.A., Ring, C., Amerongen, A.V.N., 2004. Academic examinations and immunity:
academic stress or examination stress? Psychosomatic Medicine, 625–627.
Brown, J.A., Dewey, C., Delange, C.F.M., Mandell, I.B., Purslow, P.P., Robinson, J.A.,
Squires, E.J., Widowski, T.M., 2009. Reliability of temperament tests on finishing
pigs in group-housing and comparison to social tests. Applied Animal Behaviour
Science 118, 28-35.
Burgdorf, J. & Panksepp, J., 2006. The neurobiology of positive emotions. Neuroscience &
Biobehavioral Reviews. 30, 173–187.
Burghardt, G.M., 2005. The Genesis of Animal Play: Testing the Limits. Cambridge,
Massachusetts: MIT Press.
Byers, J.A., 1977. Terrain preferences in the play behavior of Siberian ibex kids (Capra ibex
sibirica). Zeitschrift für Tierpsychologie, 45, 199-209.
Byers, J.A., Walker, C., 1995. Refining the motor play hypothesis for the evolution of play.
American Naturalist, 146, 25-40.
Cabanac, M., 1992. Pleasure: the common currency. Journal of Theoretical Biology 155, 173– 200.
107
Carreras, R, Arroyo, L., Mainau, E., Valent, D., Bassols, A., Dalmau, A., 2017. Can the way
pigs are handled alter behavioural and physiological measures of affective state?
Behavioural processes 142, 91–98.
Carver, C. S., 2001. Affect and the functional bases of behavior: on the dimensional structure
of affective experience. Personality and Social Psychology Review, 5, 345–356.
Czéh, B., Simon, M., Schmelting, B., Hiemke, C., Fuchs, E., 2006. Astroglial Plasticity in the
hippocampus is affected by psychological stress and concomitant fluoxetine treatment.
Neuropsychopharmacology, 1616–1626.
de los Campos, G., Gianola, D., Heringstad, B., 2006. A structural equation model for
describing relationships between somatic cell score and milk yield in first-lactation
dairy cows. Journal of dairy science 89. 4445–4455.
Diener, E., Lucas, R.E., 2000. Subjective emotional well-being. In M. Lewis and J.M.
Haviland-Jones (Eds.), Handbook of emotions (2nd ed., pp. 325-337). New York:
Guilford Press.
Fagen R.M., 1976. Exercise, Play, and Physical Training in Animals. In: Bateson P.P.G.,
Klopfer P.H. (eds) Perspectives in Ethology. Springer, Boston, MA.
Forkman, B., Boissy, A., Meunier-Salaün, M.-C., Canali, E., Jones, R.B., 2007. A critical
review of fear tests used on cattle, pigs, sheep, poultry and horses. Physiology &
Behavior 92 (3), 340–374.
Fraser, D., 2008. Understanding animal welfare. Acta Veterinaria Scandinavica 50.
Goursot, C., Düpjan, S., Tuchscherer, A.; Puppe, B.; Leliveld, L. M. C., 2018. Behavioural
lateralization in domestic pigs (Sus scrofa)-variations between motor functions and
individuals. In: Laterality 23 (5), 576–598.
Groffen, J., 2012. Tail posture and motion as a possible indicator of emotional state in pigs.
Animal Science program, Exam work in Animal Science EX 0567, SLU and
Wageningen University (WUR).
Hair, J.F., Hult, G.T.M., Ringle, C.M., Sarstedt, M., Richter, N.F., Hauff, S., 2017. Partial
Least Squares Strukturgleichungsmodellierung (PLS-SEM). Verlag Franz Vahlen
GmbH München.
108
Hair, J.F., Ringle, C.M., Sarstedt, M., 2011. PLS-SEM: Indeed a silver bullet, Journal of
Marketing Theory and Practice, 19, 139-151.
Hair, J.F., Tomas, G., Hult, M., Ringle, C.M., Sarstedt, M., 2014. A primer on partial least
square structural equation modeling (PLS-SEM). SAGE Publications, Inc. Thousand
Oaks, California, USA.
Hausberger, M., Fureix, C., Bourjade, M., Wessel-Robert, S., Richard-Yris, M.-A., 2012. On
the significance of adult play: what does social play tell us about adult horse welfare?
Naturwissenschaften 99, 291-302.
Henseler J., Ringle, C.M., Sinkovics, R.R., 2009. The use of partial least squares path
modeling in international marketing. Advances in International Marketing. 277–320.
Heringstad B, Wu, X.-L., Gianola, D., 2009. Inferring relationships between health and
fertility in Norwegian Red cows using recursive models. Journal of dairy science.
1778–1784.
Kiley-Worthington, M., 1975. The tail movement of ungulates, canids and felids with
particular reference to their causation and function as display. Behaviour 56: 69–115.
Lamb E, Shirtliffe, S. J., May, W. E., 2011. Structural equation modeling in the plant
sciences. An example using yield components in oat. Canadian Journal of Plant
Science 91. 603–619.
Lawrence, A., 1987. Consumer demand theory and the assessment of animal welfare. Animal
Behaviour 35, S. 293-295.
Martin, P., Caro, T.M., 1985. On the functions of play and its role in development. Advances
in the Study of Behavior. Vol. 15 (Ed. By J.L. Rosenblatt & P.J.B. Slater), 59-103.
Orlando, Florida:Academic Press.
Martin, P., Bateson, P., 2007. Measuring behaviour: an introductory guide. Cambridge
University Press. Cambridge, United Kingdom.
Müller, J., Otter, V., Lüntzel, V., Theuvsen, L., 2015. Tierwohlaspekte und die
Besuchswahrscheinlichkeit im Galopprennsport, Tierhaltung im Spannungsfeld von
Tierwohl, Ökonomie und Gesellschaft. Tagungsband zur Tierwohl-Tagung 2015 in
Göttingen, 115.
109
Müller-Schwarze, D., Stagge, B., Müller-Schwarze, C., 1982. Play behavior: persistence,
decrease, and energetic compensation during food shortage in deer fawns. Science,
215, 85-87.
Ortony, A., Turner, T.J., 1990. What’s basic about basic emotions? Psychological Review, 97,
315-331.
Reefmann, N., Bütikofer Kaszàs, F., Wechsler, B.; Gygax, L., 2009. Ear and tail postures as
indicators of emotional valence in sheep. Applied Animal Behaviour Science 118 (3-
4), 199–207.
Reimert, I., Bolhuis, J.E., Kemp, B., Rodenburg, T.B., 2013. Indicators of positive and
negative emotions and emotional contagion in pigs. Physiology and Behavior, 109,
42–50.
Ringle, C.M., 2004. Gütemaße für den Partial Least Square-Ansatz zur Bestimmung von
Kausalmodellen. Arbeitspapier Nr. 16.
Ringle, C.M., Wende, S., Becker, J.M., 2015. SmartPLS 3 (Software).
Rolls, E.T., 2005. Emotion explained. Oxford, UK: Oxford University Press.
SAS 2017. SAS Institute Inc. Cary, NC, USA.
Siviy, S.M., Harrison, K.A., Fear McGregor, L.S., 2006. Risk assessment, and playfulness in
the juvenile rat. Behavioural Neuroscience, 120 (1), S. 49-59.
Siviy, S.M., Panksepp, J., 1985. Energy-balance and play in juvenile rats. Physiology and
Behavior, 35, 435-441.
Stolba, A., Wood-Gush, D.G.M., 1980. Arousal and exploration in growing pigs in different
environments. Applied Animal Ethology, P. 382-383.
Valente B.D., Guilherme J.M.R., Gianola, D., Wu, X.-L., Weige, K., 2013. Is structural
equation modeling advantageous for the genetic improvement of multiple traits?
Genetics 194. 561–572.
110
111
GENERAL DISCUSSION
The aim of this thesis was to obtain a better understanding of livestock’s particularly positive
affective state, exemplified by investigations with fattening pigs. On the basis of an analysis
of behavioural and physiological data of fattening pigs housed in two different housing
systems from three different farms, potential indicators were to be derived which would
enable feasible and reliable measurement of the pigs’ particularly positive affective state.
According to literature, the animals’ affective state contains different emotions such as
pleasure, happiness, pain and suffering and other feelings such as hunger and thirst that are
experienced as pleasant or unpleasant (Fraser, 2008). At this point, the primarily positive
affective state presumably includes experienced pleasant emotions such as happiness (Ortony
and Turner, 1990; Diener and Lucas, 2000), whereas unpleasant emotions such as fear, pain
or suffering probably indicate the animals’ more negative affective state. Hence, in the present
thesis, the question arose of when, under what circumstances and in what manner do
domesticated fattening pigs experience pleasant emotions such as happiness that can be
paraphrased by luck, joy, contentment or cheerfulness and thereby indicate a positive
affective state. That they can feel such complex emotions seems likely as they reveal similar
brain structures and chemistry to humans (Boissy et al., 2007) and it has been already
established that animals are able to experience basic emotions such as anxiety (Panksepp,
1998) as well as more complex emotions such as tension, nervousness, calmness or
contentment (Preston and de Waal, 2002).
First, it seems quite possible that the component ‘basic health and biologic functioning’ of
good animal welfare (Fraser, 2008) must be reached satisfactorily to attain a more positive
affective state as it is probable that alarmingly sick fattening pigs do not possess a completely
positive affective state. Therefore, housing systems and treatments that prevent disease and
suffering constitute an important requirement for the pigs to possess a positive affective state.
However, actually biologic functioning in the form of performance does not necessarily
contribute to a positive affective state, but only an animal that is healthy can show high
performance, which again implies that a more positive affective state requires biologic
functioning. Moreover, it might be possible that the component ‘natural living’ of good
animal welfare (Fraser, 2008) such as social contact to conspecifics or the opportunity to
perform innate natural behavioural patterns must also be complied so that the pigs can possess
112
positive affective states. Thus, a pig may have to wallow to possess a more positive affective
state, as it is an innate behavioural trait. This becomes even more prominent due to the fact
that conventionally raised domestic pigs show the same behavioural patterns as wild boars if
they have the opportunity (Stolba and Wood-Gush, 1989). In this regard, it appears likely that
housing systems that provide opportunities for the pigs to perform these behaviours probably
induce positive affective states. Continuously, it can be assumed that housing conditions
which offer the domesticated pigs a similar environment to that of wild boars could be those
housing systems which provide the best opportunity to obtain the most positive affective state
as possible. Therewith, a basic requirement of healthy animals must be fulfilled and natural
stress factors with which wild boars are confronted (e.g. foraging, food competition, hunters,
predators or extreme climatic influences) must be reduced to a minimum. Under natural
conditions, wild boars prefer the combination of different areas of the environment such as
open and dense forests, swamps and fields (Andersson et al., 2011) where they can perform
their major activities of sleeping, resting, rooting, feeding, standing and walking (Blassetti et
al., 1988). Accordingly, in this thesis, it was hypothesised that it might be possible for the
pigs housed in the enriched habitats including different functional areas on farms 2 and 3 to
possess a more positive affective state than the pigs of the barren housing environment of
farm 1 which contained only one functional area to live out all natural behavioural patterns.
Potential reliable indicators for the assessment of the pigs’ positive affective state
A variation of behavioural tests is known to be appropriate to assess the level of fear in
animals (e.g. Murphy et al., 2014; Hemsworth and Coleman, 1998) and focus on the
assessment of emotions (Carreras et al., 2017). Therefore, a human approach (HAT) and
novel object test (NOT) were selected to be applied to the pigs (Chapter One). As the
occurrence of playing behaviour is said to exist when animals experience positive emotions
(Fraser and Duncan, 1998; Špinka et al., 2001; Barnard, 2004; Burgdorf and Panksepp, 2006)
the pigs’ playing behaviour was also examined as a potential indicator to measure the pigs’
positive affective state (Chapter Two). In addition, certain body language signals such as the
pigs’ tail postures are described to vary depending on the emotional context (Reimert et al.,
2013) and therefore the occurrence of different tail and also ear postures (curled-up, hanging,
wagging, raised and tucked-under tails respectively ears directed forwards, backwards,
laterally and mixed) were investigated as well (Chapter Three). Further, the partial least
squares structural equation modelling method was used to identify the affective state’s
113
influence on the parameters under observation as well as on further physiological parameters
and the potential suitability to measure the fattening pigs’ positive affective state of these
parameters (Chapter Four).
Behavioural tests
Especially at the end of fattening, the results of Chapter One showed consistent results
between the two different housing systems of the approach latencies in both behavioural tests,
which probably indicate the pigs’ more positive affective state in the enriched housing system
than the barren one. Higher approach latencies to come into contact with novel stimuli, such
as unknown humans or novel objects, are associated with more fearful animals (Brown et al.,
2009) and thus with a less positive affective state. However, the higher approach latencies
could also be a reason for less motivation to explore novel stimuli (Stolba and Wood-Gush,
1980) related to less boredom due to the enriched environment (Meagher and Mason, 2012),
which is then again an indicator of a more positive affective state. Simultaneously, barren-
housed pigs might be more motivated to explore a novel object or a person in the home pen
probably due to fewer exploration possibilities in their habitat compared to enriched-housed
pigs, which could explain the quicker approach latencies of the barren-housed pigs in this
study and presumably identify their less positive affective state. However, Forkman et al.
(2007) claim that the avoidance reaction of the animal appears to be essential as both a non-
curious and a fearful animal shows long latencies to approach, as mentioned above. It has
been also stated that the animals’ responsiveness towards humans and novel objects could be
considered to be rather an individual temperament trait (Gibbons et al., 2009) and not
dependent on the affective state. Therefore, it might be difficult to use the approach latencies
of HATs and NOTs as the sole indicators to measure the pigs’ positive affective state.
Including avoidance reaction or body language signals, such as the tail or ear postures during
the test situation, might enhance the possible informative power of these behavioural tests.
Furthermore, the implementation of a forced human approach test could be helpful (Pedersen
et al., 2003) in detecting the factual reason for the avoidance reaction (little interest or fear).
The forced human approach test might increase the likelihood of an animal responding more
actively to a human, whereas in the voluntary human approach test the chances of getting no
response or a passive response might probably be higher. Also the implementation of a forced
novel object test could be conceivable, which has also the advantage of minimising the impact
of the person who performs the behavioural test. Previous studies have shown that the pigs’
114
behaviour in HATs can be influenced by their previous experiences with humans as there are
studies which demonstrate that negative handling leads to more avoidance responses (Carreras
et al., 2017), whereas positive handling might reduce the approach latency (Büttner et al.,
2018). Here, also higher approach latencies of male rather than female pigs in a HAT can be
explained by their castration and its consequences related to humans (Reimert et al., 2013a).
Additionally, the individual distance of each pig could also influence its behaviour in the
HAT (Rutherford et al., 2012) so that it is not necessarily fear that explains a high approach
latency but rather the character of the animal. Hence, it could be conceivable that the pigs’
behaviour in the NOT is better suited than the pigs’ behaviour in the HAT to estimate the
pigs’ positive affective state, which is confirmed by the indicator reliabilities of Chapter
Four. Additionally, the note of the pigs’ defecation and urination during the test phases could
be valuable as they could stand for negative experienced emotions in such test situations
(Forkman et al., 2007; Reimert et al., 2014). More elaborate in the application, but also useful
to receive more detailed information of the pigs’ affective state, could be the implementation
of cognitive bias tests. Here, animals with positive affective states behave also more
positively in new test situations compared to animals with rather negative affective states,
which can reveal cognitive bias tests as suitable indicators to assess affective states (Harding
et al., 2004). Furthermore, it could be investigated whether pigs which have been previously
positively conditioned to humans or novel objects approach them for instance with a curled-
up tail. This could combine the suitability of behavioural tests and body language signals as
an indication of the pigs’ affective state.
Another important consideration in further studies should be that all pens in which the pigs
are tested comprise the same size and design, which can possibly influence the pigs’
behaviour during the test situations (e.g. no hiding or feeding places). Here, it should be noted
that also the use of a home pen or a test pen and the test procedure at group or individual level
in behavioural tests could influence the tested pigs’ behaviour. For instance, an animal which
is tested alone could possibly show fear responses due to the effect of social isolation
(Forkman et al., 2007). Additionally, a test procedure at group level could probably hamper
the knowledge gained for each individual pig and also the habituation process in a test pen
could affect further behavioural responses (Forkman et al., 2007). Thus, when interpreting
results, it is important to consider how the behavioural tests were conducted, as each provides
advantages and disadvantages as described.
115
Playing behaviour
Particularly at the end of fattening, the enriched-housed pigs showed longer total durations of
playing behaviour and solely locomotor play than the barren-housed pigs (Chapter Two).
Thus, it appears realistic that playing behaviour could depict a suitable indicator to measure
the primarily positive affective state in pigs as, for instance, it has been stated that animals
only show playing behaviour when optimal environmental conditions prevail (Lawrence,
1987; Held and Špinka, 2011) and they fare well and feel well (Siviy et al., 2006). Hence,
playing behaviour is known as “luxury” behaviour (Lawrence, 1987), which seems to
decrease when animals experience negative emotions (Fagen 1976; Martin and Caro 1985;
Burghardt, 2005) as it has been often described as being connected with animals experiencing
positive emotions (Fraser and Duncan, 1998; Špinka et al., 2001; Barnard, 2004; Burgdorf
and Panksepp, 2006). This confirms the assumption that playing behaviour represents a
reliable indication of a positive affective state, which is in accordance with the results of the
indicator reliabilities in Chapter Four. Mainly the enrichment of the environment (Špinka et
al., 2001) and increased availability of space could explain the longer occurrence of playing
pigs in the enriched housing systems compared to the barren environment in this study. Here,
further research studies should investigate whether an increased availability of space in barren
habitats also augments the occurrence of playing behaviour or whether it is primarily the
enrichment of the environment such as straw-bedding which causes the pigs to show more
playing behaviour. Here, the ground may also be an important factor which influences
whether the pigs show more playing behaviour as a slippery slatted floor probably does not
encourage the pigs to perform playing behaviour. The increased availability of space
presumably also enables more locomotor play of the enriched-housed pigs compared to the
barren-housed pigs, which is even more suitable than the total duration of playing behaviour
to assess the pigs’ positive affective state, as demonstrated by the structural equation model in
Chapter Four. This could be because the total duration of playing behaviour includes, apart
from locomotor play, also the durations of social play, which is possibly less suitable to
indicate the pigs’ especially positive affective state due to controversially discussed literature.
For instance, Hausberger et al. (2012) examined the social play behaviour of captive horses
and their physiological parameters such as vertebral disorders and linked significantly
increased occurrence of social play to poor welfare. In this study, socially playing horses
showed more vertebral disorders and behaved more aggressively towards humans than non-
socially playing horses. Thus, social play possibly indicates an opportunity for animals to
116
cope with their usual, unfavourable life conditions (Hausberger et al., 2012) and reveals a
negative rather than positive affective state. In addition, it can be difficult to distinguish the
pigs’ social play from real fighting behaviour. Considering this, in further studies,
investigations of object play could be more suitable to obtain reliable information on the pigs’
primarily positive affective state, but it was not possible to compare this in the present thesis
as both housing systems did not provide the same opportunities for the pigs to perform object
play. Object play is defined by behavioural patterns when the animals manipulate, shake or
carry objects around (Newberry et al., 1988) and presumably is less impacted by a correct
detection or pen mates than social play and therefore possibly better suited than social play to
identify the pigs’ exclusively positive affective state more reliably. However, it might be
difficult to correctly distinguish object play from exploration behaviour, but this would less
affect the reliable detection of the positive affective state since the opportunity to perform
exploratory behaviour constitutes an important part of good animal welfare (Wood-Gush and
Vestergaard, 1990). Another important consideration in further studies should be to have the
same opportunities to record the required video data on all tested farms; so that a more
standardised analysis of the pigs’ performed playing behaviour is feasible. This could avoid
possible impact factors such as different daylight-length or visual qualities of the video data
which could influence the analysed playing behaviour. In addition, in future studies, it would
be preferable to examine the pigs’ playing behaviour on experimental farms including
completely standardised conditions, since also the management system – whether the farms
operate in a closed system or represent a pure fattening farm – can have impacts on the pigs’
behaviour. For instance, a relative deterioration of the familiar habitat due to another possibly
less attractive housing system could result in frustrated pigs which do not feel well and
therefore show less playing behaviour (Siviy et al., 2006; Le Floc'h et al., 2010).
Body language signals
Particularly at the end of fattening, the enriched-housed pigs showed more curled-up tails than
the barren-housed pigs (Chapter Three). This is in accordance with other studies in which
enriched-housed pigs indicated more often a curled-up tail than barren-housed pigs (Groffen,
2012). While there are usually more possibilities in enriched habitats to explore than in barren
housing systems, the more often observed curled-up tails in the enriched housing system
could be due to the satisfied need of the pigs to perform their natural living (Fraser, 2008).
This ability to perform natural behaviour and a simultaneous increase in the occurrence of
117
curled-up tails could indicate that pigs when they curl up their tails are satisfied and therefore
presumably possess a more positive affective state. Pigs with curled-up tails were also
observed in rewarding situations (Goursot et al., 2018), which can be considered as positive
experiences and therefore possibly indicate a more positive affective state as well. Moreover,
authors observed a positive correlation between feeding and drinking with curled-up tails,
which can also be considered a positive experience and, therewith, presumably indicate the
pigs’ more positive affective state. Thus, it seems acceptable that curled-up tails could be
potential indicators to identify the pigs’ positive affective state. Here, the results of Chapter
Four confirm this assumption. However, a key point to improve the experimental set-up
should be to combine the notification of the tail and ear postures with the related performed
activity at the point of scan-sampling as those presumably influence their meaning. For
instance, wagging tails were also observed during feeding (Kleinbeck and McGlone, 1993),
which are discussed controversially as an indication of an exclusively positive affective state.
In general, it may be difficult to assess the tail and ear postures during the feeding of pigs,
since, for instance, the animal feeding place ratio could also influence their meaning.
Moreover, regarding the predominant pig production conditions of intensive housing systems
in European countries in which the pigs’ tails are usually docked to protect them from injuries
by tail-biting, it is difficult to apply a curled-up tail as an indication of a more positive
affective state because docked tails cannot be curled up. Thus, only in housing systems where
the pigs’ tails are undocked could a curled-up tail be used to identify a more positive affective
state. Hence, deliberations could arise that pigs when they are not able to curl up their tails
can possibly possess a less positive affective state compared to pigs whose tails are undocked.
However, tail-docking is conducted on many farms with prior approval to protect the pigs
from injury and suffering caused by tail-biting, which presumably increases the positive
affective state rather than decreases it. Nevertheless, this deliberation seems also less realistic
as curled-up tails do not occur in wild boars (Jensen, 2002) although it could be conceivable
that wild boars, which seem to be able to perform their natural living almost unrestricted,
must show a curled-up tail in general. However, this non-appearance of curled-up tails in wild
boars could be explained by the fact that the pigs’ ability to curl up the tail developed during
the domestication process (Goursot et al., 2018).
The investigations of Chapter Four also indicated the ears directed laterally as suitable
indicator to assess the pigs’ positive affective state. However, this finding remains
questionable, as it seems imaginable that the ears directed laterally of the study in Chapters
118
Three respectively Four could be related to both a passive ear posture (ears hanging down,
relaxed) or a change in ear posture. Reimert et al. (2013) observed more ear posture changes
in pigs in aversive situations such as social isolation combined with negative, unpredictable
interventions, which could be an indication of a more negative affective state. In contrast to
this, higher amounts of passive ear postures in sheep occurred more in positive than in
negative situations (Reefmann et al., 2009). Though it should be noted that due to the
snapshot by the scan-sampling method of the video data in the study of Chapter Three, it
cannot be unequivocally evaluated whether the ears directed laterally are related to ear posture
changes or rather passive ear postures and both presumably could be able to indicate negative
respectively positive affective states. Conclusively, it seems difficult to conclude clearly
whether the ears directed laterally can be due to the pigs’ more positive or rather negative
affective state, although the results of Chapter Four indicate this ear posture as a suitable
indicator. In further studies, this could be improved with more cameras from multiple
perspectives of the pen or the use of continuous sampling. With these improvements, it should
be easier to identify which position the ears need to be assigned to.
Conclusion and outlook
This thesis has shown that the approach latencies of both behavioural tests, total duration of
playing behaviour and locomotor play as well as curled-up tails could probably constitute
reliable indicators to identify the pigs’ particularly positive affective state. These indicators
revealed consistent results between both examined housing systems which are in accordance
with previous literature and can therefore draw conclusions concerning the pigs’ affective
state. Nevertheless, until now, primarily locomotor play and total duration of playing
behaviour as well as the curled-up tails have seemed to be reliable to measure the pigs’
positive affective state as the sole indicators. The approach latencies of both behavioural tests
in combination with the previously mentioned parameters appeared to be reliable to assess the
pigs’ particularly positive affective state. However, it is conceivable that these parameters,
especially the curled-up tails, could be integrated as feasible indicators into existing animal
welfare assessment systems to obtain a better understanding of the pigs’ positive affective
state.
In future studies, behavioural parameters such as e.g. the forced human-approach or forced
novel-object tests and cognitive bias tests should be further investigated regarding their
suitability to measure the pigs’ positive affective state. These parameters could be useful to
119
compensate the existing interpretation difficulties regarding high approach latencies and could
therefore capture the pigs’ positive affective state using behavioural tests as sole indicators
more reliably and conceivably complement the reliable suitability of other behavioural
parameters.
In closing, the conclusions presented in this thesis lay relevant foundations for the
understanding and measurement of the fattening pigs’ positive affective state. Further
research should supplement the studies of this thesis to obtain a better understanding of the
affective state in other animal species as well.
120
REFERENCES
Andersson, A., Äänismaa, R., Huusko, J., Jensen, P., 2011. Behaviour of European wild boars
(Sus scrofa) in connection with farrowing in an enclosure. Mammalian Biology, (76),
3, 332-338.
Barnard, C. J., 2004. Mechanism, Development, Function and Evolution. Animal Behaviour,
Upper Saddle River, New Jersey: Pearson Education.
Blasetti, A., Boitani, L., Riviello, M.C., Visalberghi, E., 1988. Activity Budgets and Use of
Enclosed Space by Wild Boars (Sus scrofa) in Captivity. Zoo Biology 7, 69-79.
Boissy, A., Manteuffel, G., Jensen, M. B. Moe, R. O., Spruijt, B., Keeling, L.J., Winckler, C.,
Forkman, B., Dimitrov, I., Langbein, J., Bakken, M., Veissier, I., Aubert, A., 2007.
Assessment of positive emotions in animals to improve their welfare. Physiology and
behavior 92, 375–397.
Brown, J.A., Dewey, C., Delange, C.F.M., Mandell, I.B., Purslow, P.P., Robinson, J.A.,
Squires, E.J., Widowski, T.M., 2009. Reliability of temperament tests on finishing
pigs in group-housing and comparison to social tests. Applied Animal Behaviour
Science 118, 28-35.
Burgdorf, J., Panksepp, J., 2006. The neurobiology of positive emotions. Neuroscience and
Biobehavioral Reviews, 30, 173-187.
Burghardt, G.M., 2005. The Genesis of Animal Play: Testing the Limits. Cambridge,
Massachusetts: MIT Press.
Büttner, K., Czycholl, I., Basler, H., Krieter, J., 2018. Effects of an intensified human-animal
interaction on tail-biting in pigs during the rearing period. The Journal of Agricultural
Science, 1-8.
Carreras, R, Arroyo, L., Mainau, E., Valent, D., Bassols, A., Dalmau, A., 2017. Can the way
pigs are handled alter behavioural and physiological measures of affective state?
Behavioural processes 142, 91–98.
Diener, E., Lucas, R.E., 2000. Subjective emotional well-being. In M. Lewis and J.M.
Haviland-Jones (Eds.), Handbook of emotions (2nd ed., pp. 325-337). New York:
Guilford Press.
121
Fagen, R. M., 1976. Exercise, Play, and Physical Training in animals. In: Bateson P.P.G.,
Klopfer P.H. (eds) Perspectives in Ethology. Springer, Boston, MA.
Forkman, B., Boissy, A., Meunier-Salaün, M.-C., Canali, E., Jones, R. B., 2007. A critical
review of fear tests used on cattle, pigs, sheep, poultry and horses. Physiology and
Behavior 92 (3), 340–374.
Fraser, D., 2008. Understanding animal welfare. Acta Veterinaria Scandinavica 50.
Fraser, D., Duncan, I. J. H., 1998. ‘Pleasures’,‘pains’ and animal welfare: toward a natural
history of affect. Animal Welfare, 7, 383-396.
Gibbons, J. Lawrence, A., Haskell, M., 2009. Responsiveness of dairy cows to human
approach and novel stimuli. Applied Animal Behaviour Science 116, 163-173.
Goursot, C., Düpjan, S., Tuchscherer, A., Puppe, B., Leliveld, L.M.C., 2018. Behavioural
lateralization in domestic pigs (Sus scrofa)-variations between motor functions and
individuals. In: Laterality 23 (5), 576–598.
Groffen, J., 2012. Tail posture and motion as a possible indicator of emotional state in pigs.
Animal Science program, Exam work in Animal Science EX 0567, SLU and
Wageningen University (WUR).
Harding, E.J., Paul, S., Mendl, M., 2004. Cognitive bias and affective state. Nature 427, 312.
Hausberger, M., Fureix, C., Bourjade, M., Wessel-Robert, S., Richard-Yris, M.-A., 2012. On
the significance of adult play: what does social play tell us about adult horse welfare?
Naturwissenschaften 99, 291-302.
Held, S.D.E., Špinka, M., 2011. Animal Play and animal welfare. Animal Behaviour 81, 891-
899.
Hemsworth, P.H., Coleman, G.J., 1998. Human–Livestock Interactions: The Stockperson and
the Productivity of Intensively Farmed Animals. CAB International, Wallingford, UK.
Jensen, P., 2002. The ethology of domestic animals: An introductory text. New York: CABI.
Kleinbeck, S., McGlone, J.J., 1993. Pig tail posture: A measure of stress. Technical
University Agricultural Science Technical Report T-5-327: 47–48.
122
Lawrence, A., 1987. Consumer demand theory and the assessment of animal welfare. Animal
Behaviour 35, S. 293-295.
Le Floc'h, N., Matte, J. J., Melchior, D., van Milgen, J., Sève, B., 2010. A moderate
inflammation caused by the deterioration of housing conditions modifies Trp
metabolism but not Trp requirement for growth of post-weaned piglets. Animal: an
international journal of animal bioscience 4 (11), 1891–1898.
Martin, P., Caro, T.M., 1985. On the functions of play and its role in development. Advances
in the Study of Behavior. Vol. 15 (Ed. By J.L. Rosenblatt & P.J.B. Slater), 59-103.
Orlando, Florida:Academic Press.
Meagher, R.K., Mason, G.J., 2012. Environmental Enrichment Reduces Signs of Boredom in
Caged Mink. Environmental Enrichment Reduces Signs of Boredom in Caged Mink.
PLoS ONE 7(11): e49180.
Murphy, E., Nordquist, R.E., van der Staay, F.J., 2014. A review of behavioural methods to
study emotion and mood in pigs, Sus scrofa. Applied Animal Behaviour Science. P. 9-
28.
Newberry, R.C., Wood-Gush, D.G.M., Hall, J.W., 1988. Playful behaviour of piglets.
Behavioural Processes 17 (3), 205–216.
Ortony, A., Turner, T.J., 1990. What’s basic about basic emotions? Psychological Review, 97,
315-331.
Panksepp, J., 1998. Affective neuroscience: The foundations of human and animal emotion,
New York: Oxford University Press.
Pedersen, L.J., Damm, B.I., Kongsted, A.G., 2003. The influence of adverse or gentle
handling procedures on sexual behaviour in fearful and confident sows. Applied
Animal Behaviour Science 83, 277-290.
Preston, S.D., de Waal, F.B.M., 2002. The communication of emotions and the possibility of
empathy in animals, In: Post SG, Underwood LG, Schloss JP and Hurlbut W, B. (eds)
Altruism and altruistic love, 284-308. Oxford University Press: New York.
123
Reefmann, N., Bütikofer Kaszàs, F., Wechsler, B., Gygax, L., 2009. Ear and tail postures as
indicators of emotional valence in sheep. In: Applied Animal Behaviour Science 118
(3-4), 199–207.
Reimert, I., Bolhuis, J.E., Kemp, B., Rodenburg, T.B., 2013. Indicators of positive and
negative emotions and emotional contagion in pigs. Physiology and Behavior, 109,
42–50.
Reimert, I., Rodenburg, T. B., Ursinus, W. W., Kemp, B.; Bolhuis, J. E., 2014. Responses to
novel situations of female and castrated male pigs with divergent social breeding
values and different backtest classifications in barren and straw-enriched housing.
Applied Animal Behaviour Science 151, 24–35.
Reimert, I., Rodenburg, T.B., Ursinus, W.W., Duijvesteijn, N., Camerlink, I., Kemp, B.,
Bolhuis, J.E., 2013a. Backtest and novelty behavior of female and castrated male
piglets with diverging social breeding values for growth. Journal of Animal Science
91, 4589-4597.
Rutherford, K.M.D., Donald, R.D., Lawrence, A.B., Wemelsfelder, F., 2012. Qualitative
Behavioural Assessment of emotionality in pigs. Applied Animal Behaviour Science
139, 218-224.
Siviy, S.M., Harrison, K.A., McGregor, I.S., 2006. Fear, risk assessment, and playfulness in
the juvenile rat. Behavioural Neuroscience, 120 (1), 49-59.
Špinka, M., Newberry, R. C. & Bekoff, M., 2001. Mammalian play: training for the
unexpected. Quarterly Review of Biology, 76, 141-168.
Stolba, A., Wood-Gush, D.G.M., 1980. Arousal and exploration in growing pigs in different
environments. Applied Animal Ethology, 382-383.
Stolba, A., Wood-Gush, D.G.M., 1989. The behaviour of pigs in a semi-natural environment.
In: Animal Science 48 (2), 419-425.
Wood-Gush, D. G. M. & Vestergaard, K., 1990. Exploratory behaviour and the welfare of
intensively kept animals. Journal of Agricultural and Environmental Ethics, 2, 161-
169.
124
125
GENERAL SUMMARY
The aim of this thesis was to derive potential indicators to attain a reliable measurement and
better understanding of livestock’s particularly positive affective state, exemplified by
investigations with fattening pigs. Diverse behavioural and physiological parameters were
examined regarding their suitability to assess the pigs’ primarily positive affective state in two
different housing systems on three different farms which differed especially in terms of an
enriched or barren environment, availability of space and climatic conditions.
In Chapter One, behavioural tests such as the human approach (HAT) and novel object test
(NOT) were investigated with regard to their usefulness to assess the fattening pigs’ positive
affective state. The pigs (n=297) were subjected to behavioural tests three times during their
fattening (begin, middle and end of fattening). They were tested alone in their home pen for
three minutes whereby three variables were analysed: the approach latency, duration of
contacts and number of contacts. Especially at the end of fattening, the enriched-housed pigs
showed higher approach latencies than the barren-housed pigs in both behavioural tests. The
HAT indicated lower durations of contacts in the enriched- than barren-housed pigs in
particular at the end of fattening but no clear differences between the two housing systems in
the NOT. For the number of contacts, there were no clear differences between the two
housing systems in both behavioural tests. The higher approach latencies could be due to less
motivation to explore novel stimuli related to an enriched environment with sufficient
opportunities to perform natural living. As natural living constitutes an important part of
animal welfare, the finding of the higher approach latencies in the enriched environment
might point to a more positive affective state of these pigs and could therefore represent a
potential indication of the same.
Chapter Two examined the occurrence of the pigs’ (n=228) total duration of playing
behaviour (durations of locomotor and social play in total), locomotor play and social play in
the two different housing systems. Video data from two days at the beginning and two days at
the end of fattening were analysed to obtain the durations of the different playing behaviours
during the day length (s/h) by the use of the continuous-sampling method. At the beginning of
fattening, the barren-housed pigs showed longer total durations of playing behaviour than the
enriched-housed pigs. At the end of fattening, an inverse situation existed as the enriched-
housed pigs showed longer total durations of playing behaviour than the barren-housed pigs.
126
The locomotor play did not indicate clear differences between the two housing systems at the
beginning of fattening, but more locomotor play occurred in the enriched-housed pigs than the
barren-housed ones at the end of fattening. In contrast, at the beginning of fattening, the
barren-housed pigs showed more social play than the enriched-housed pigs and no clear
differences occurred between the two housing systems at the end of fattening. To conclude,
especially the consistent results of the total playing behaviour and locomotor play between the
two housing systems appeared to be suitable to determine the pigs’ positive as well as
negative affective states. This can be explained by the assumption that playing behaviour
increases when animals experience positive emotions and decreases when they are affected by
negative conditions. Due to controversially discussed literature and inconclusive results
especially at the end of fattening, social play did not reveal itself to be appropriate to assess
the pigs’ positive affective state.
In Chapter Three, the body language signals of the pigs (n=228) were assessed regarding the
occurrence of the different tail and ear postures (curled-up, hanging, raised, tucked-under or
wagging tails respectively ears directed forwards, backwards, mixed and laterally) in the two
different housing systems. Two days of the beginning and two days of the end of fattening
were analysed by using the scan-sampling method. Particularly at the end of fattening, the
enriched-housed pigs showed more curled-up tails than the barren-housed pigs. The barren-
housed pigs showed also more raised and wagging tails than the enriched-housed pigs.
Especially at the end of fattening were no clear differences of the ears directed forwards
between the two housing systems and the barren-housed pigs indicated fewer ears directed
laterally than the enriched-housed pigs. In this study, primarily the curled-up tails appeared to
be appropriate to assess the pigs’ positive affective state due to consistent results between the
two different housing systems and the assumption that the occurrence of curled-up tails is
related to satisfying situations. All other tail respectively ear postures were interpreted to be
less reliable to measure the pigs’ positive affective state related to their rare occurrence,
controversially discussed literature or inconsistent results between the two different housing
systems.
Chapter Four evaluated the relationships between the large variety of parameters for
behaviour (behavioural tests, playing behaviour and body language signals) and physiology
(diameters and astroglia cell numbers of hippocampi, salivary immunoglobulin-A content
respectively protein compositions) and the pigs’ positive affective state as well as its influence
127
on the parameters mentioned. The partial least squares structural equation modelling method
was applied to identify these relationships. The assessment of the coefficients of
determination (R²) evaluated the variables of the behavioural tests, body language signals and
playing behaviour to be influenced by the pigs’ positive affective state. The indicator
reliabilities revealed locomotor play, ears directed laterally, curled-up tails, playing behaviour
and approach latencies of both behavioural tests, in this order, to be the most appropriate
parameters to estimate the latent variable of the fattening pigs’ primarily positive affective
state. These findings are in accordance with the interpretations of the previous chapters and
confirm that the approach latencies of both behavioural tests, playing behaviour, curled-up
tails and especially locomotor play are potential indicators to measure the pigs’ positive
affective state reliably. Particularly the physiological parameters of this study appeared to be
less suitable to estimate the pigs’ positive affective state.
In summary, this thesis examined different methods and indicators with regard to their
suitability to assess the pigs’ positive affective state. Therewith, this thesis lays relevant
foundations to measure the affective state and contributes to its understanding in general.
128
129
ZUSAMMENFASSUNG
Das Ziel der vorliegenden Arbeit war die Identifizierung potentiell geeigneter Indikatoren, um
ein besseres Verständnis und eine zuverlässige Erfassung des positiven Gemütszustandes von
Nutztieren zu erlangen, wofür exemplarische Untersuchungen an Mastschweinen
durchgeführt wurden. Dazu wurden verschiedene verhaltensbezogene und physiologische
Parameter hinsichtlich ihrer Eignung einen positiven Gemütszustand der Schweine zu
identifizieren überprüft. Die untersuchten Mastschweine wurden auf drei Betrieben mit zwei
verschiedenen Haltungssystemen gehalten, wobei sich diese vor allem bezüglich der
reizärmeren oder reizvolleren Haltungsumgebung, des Platzangebotes pro Tier und der
klimatischen Bedingungen unterschieden.
Im ersten Kapitel wurde untersucht, ob der human approach (HAT) und novel object test
(NOT) dafür geeignet sind, einen positiven Gemütszustand der untersuchten Mastschweine
(n=297) zu erfassen. Die Verhaltenstests wurden mit jedem Tier an drei Zeitpunkten während
des Mastverlaufes durchgeführt (Vor-, Mittel- und Endmast). Hierbei wurden die Schweine
einzeln drei Minuten lang in ihrer gewohnten Bucht getestet, wobei drei verschiedene
Variablen analysiert wurden: die Annäherungslatenzzeiten, die Gesamtkontaktdauern und die
Anzahl der Kontakte. Insbesondere während der Endmast, zeigten sich in beiden
Verhaltenstests in der reizvolleren Haltungsumgebung höhere Annäherungslatenzzeiten, als in
der reizärmeren Haltungsumgebung. Zudem identifizierte der HAT vor allem in der Endmast
niedrigere Gesamtkontaktdauern der Mastschweine in der reizvolleren, als in der reizärmeren
Haltung. Im NOT zeigten sich hinsichtlich der Gesamtkontaktdauern jedoch keine deutlichen
Unterschiede zwischen den beiden Haltungssystemen. Bezüglich der Anzahl der Kontakte
entstanden keine eindeutigen Unterschiede zwischen den beiden Haltungsumgebungen in
beiden Verhaltenstests. Die höheren Annäherungslatenzzeiten könnten durch eine geringere
Motivation neue Stimuli zu erkunden bedingt sein, welche folglich im Zusammenhang mit
einer reizvolleren Haltungsumgebung und demnach ausreichenden Möglichkeiten natürliche
Verhaltensweisen auszuführen stehen. Da das Ausleben natürlichen Verhaltens einen
wichtigen Teil von Tierwohl darstellt, könnten die höheren Annäherungslatenzzeiten in der
reizvolleren Haltungsumgebung auf einen positiveren Gemütszustand dieser Mastschweine
hindeuten und gleichzeitig potentielle Indikatoren für diesen darstellen.
130
Das zweite Kapitel fokussierte sich auf das Auftreten der Gesamtdauer des Spielverhaltens
(Gesamtdauer des lokomotorischen und sozialen Spielverhaltens) und des lokomotorischen
sowie sozialen Spielverhaltens der untersuchten Mastschweine (n=228) aus den zwei
unterschiedlichen Haltungssystemen. Hierbei wurde das Videomaterial von jeweils zwei
Tagen der Vor- und Endmast mithilfe der „continuous-sampling“ Methode analysiert, um die
Gesamtdauern des unterschiedlichen Spielverhaltens (s/h) zu erhalten. Während der Vormast
zeigten die Mastschweine der reizärmeren Haltungsumgebung eine höhere Gesamtdauer des
Spielverhaltens als die Tiere des reizvolleren Haltungssystems. Hingegen änderte sich dieses
Verhaltensmuster im Verlauf der Mast, sodass zum Zeitpunkt der Endmast in der reizvolleren
Haltungsumgebung höhere Gesamtdauern des Spielverhaltens auftraten, als in der
reizärmeren Haltungsumgebung. Bezüglich des lokomotorischen Spielverhaltens zeigten sich
während der Vormast keine eindeutigen Unterschiede zwischen den beiden
Haltungssystemen, wobei jedoch während der Endmast in der reizvolleren Haltungsumwelt
mehr lokomotorisches Spielverhalten gezeigt wurde, als in dem reizärmeren Haltungssystem.
Hinsichtlich des sozialen Spielverhaltens zeigten die Mastschweine der reizärmeren
Haltungsumgebung während der Vormast mehr soziales Spiel als die Tiere der reizvolleren
Haltungsumwelt, wobei jedoch während der Endmast keine deutlichen Unterschiede zwischen
den beiden Haltungssystemen auftraten. Schließlich könnten insbesondere die eindeutigen
Unterschiede der Gesamtdauer des Spielverhaltens und des lokomotorischen Spiels zwischen
den beiden Haltungssystemen geeignet sein, sowohl einen positiveren, als auch einen
negativeren Gemütszustand der untersuchten Mastschweine zu identifizieren. Diese Annahme
ergibt sich durch die Tatsache, dass Spielverhalten vermehrt auftritt, wenn Tiere positive
Emotionen erfahren und es sich verringert, wenn sie durch negative Bedingungen beeinflusst
werden. Hingegen zeigte sich das soziale Spielverhalten aufgrund von kontrovers diskutierter
Literatur und widersprüchlichen Ergebnissen zwischen den beiden Haltungssystemen
während der Endmast als weniger geeignet, um einen positiven Gemütszustand der
untersuchten Mastschweine zuverlässig zu erfassen.
Im dritten Kapitel wurden die körpersprachlichen Signale der Mastschweine (n=228)
hinsichtlich des Auftretens von verschiedenen Schwanz- und Ohrhaltungen (geringelte,
hängende, erhobene, eingeklemmte oder wedelnde Schwänze bzw. nach vorne, hinten,
gemischt und seitlich gerichteten Ohren) in den beiden Haltungssystemen analysiert. Hierfür
wurde das Videomaterial von jeweils zwei Tagen der Vor- und zwei Tagen der Endmast
mittels der „scan-sampling“ Methode untersucht. Vor allem während der Endmast wurden in
131
der reizvolleren Haltungsumgebung mehr geringelte Schwänze beobachtet, als in der
reizärmeren Haltungsumwelt. Zudem zeigten die Schweine der reizärmeren Haltung mehr
erhobene und wedelnde Schwänze, als die Tiere des reizvolleren Haltungssystems. Außerdem
wurden insbesondere in der Endmast keine eindeutigen Unterschiede der nach vorne
gerichteten Ohren zwischen den beiden Haltungssystemen beobachtet und die Mastschweine
der reizärmeren Haltungsumgebung zeigten weniger seitlich gerichtete Ohren, als die Tiere
der reizvolleren Haltungsumwelt. In dieser Studie wurden vor allem die geringelten Schwänze
als geeignet interpretiert, um einen positiven Gemütszustand der untersuchten Mastschweine
zuverlässig zu erfassen. Dieses wurde insbesondere aufgrund von eindeutigen Ergebnissen
zwischen den beiden Haltungssystemen und der Annahme, dass geringelte Schwänze
vermehrt in zufriedenstellenden Situationen auftreten, geschlussfolgert. Die übrigen Schwanz-
und Ohrhaltungen wurden infolge von seltener Beobachtung, kontrovers diskutierter Literatur
oder widersprüchlichen Ergebnissen zwischen den beiden Haltungssystemen als weniger
geeignet identifiziert, um einen positiven Gemütszustand von Mastschweinen zu
identifizieren.
Das vierte Kapitel untersuchte die Zusammenhänge zwischen den verschiedenen
verhaltensbezogenen (Verhaltenstests, Spielverhalten und körpersprachlichen Signale) und
physiologischen (Durchmesser und Astrogliazellzahl von Hippocampi und Speichel-
Immunglobulin A Gehalt bzw. -proteinzusammensetzung) Parametern und dem positiven
Gemütszustand der Mastschweine sowie dessen Einfluss auf die genannten Parameter. Die
Methode der Strukturgleichungsmodellierung wurde angewendet, um diese Zusammenhänge
zu identifizieren. Hierbei zeigte die Bewertung der Determinationskoeffizienten (R²) einen
Einfluss des positiven Gemütszustandes von Mastschweinen auf die Verhaltenstests,
körpersprachlichen Signale und das Spielverhalten. Die Indikator Reliabilitäten identifizierten
das lokomotorische Spielverhalten, die seitlich gerichteten Ohren, die geringelten Schwänze,
die Gesamtdauer des Spielverhaltens und die Annäherungslatenzzeiten der Verhaltenstests in
dieser Reihenfolge als die geeignetsten Parameter um die latente Variable des positiven
Gemütszustandes von Mastschweinen zu schätzen. Diese Erkenntnisse entsprechen den
Interpretationen der vorherigen Kapitel und bestätigen die vermutete Eignung der
Annäherungslatenzzeiten der beiden Verhaltenstests, der Gesamtdauer des Spielverhaltens,
der geringelten Schwänze und insbesondere des lokomotorischen Spielverhaltens als
potentielle Indikatoren, um einen positiven Gemütszustand von Mastschweinen zuverlässig zu
132
messen. Vor allem die physiologischen Parameter dieser Studie schienen weniger geeignet,
den positiven Gemütszustand der untersuchten Mastschweine zu erfassen.
Insgesamt wurden in allen Kapiteln dieser Arbeit verschiedene Methoden und Indikatoren
hinsichtlich ihrer zuverlässigen Eignung für die Beurteilung des positiven affektiven Zustands
von Mastschweinen untersucht. Somit legt diese Arbeit relevante Grundlagen für die Messung
des affektiven Zustands und trägt zu dessen allgemeinem Verständnis bei.
133
ATTACHMENT
Material and Methods
The following sections contain a detailed summary of the material and methods used in the
present thesis.
Data collection
Data collection was conducted over a period from November 2016 until September 2017 in
two different housing systems on three farms in Schleswig Holstein and Lower Saxony,
Northern Germany. In the entire study, 302 crossbreed fattening pigs (Pietrain x (Large White
x Landrace)) from two batches were tested in total. The tails of the fattening pigs were
undocked, and the boars were surgically castrated. The main characteristics of the two
housing systems respectively three different farms are shown in Table 1. The housing systems
differed primarily in terms of space availability (m²/pig), an enriched or barren habitat and
climatic conditions.
Table 1: Overview of the farms included in the present thesis (LW=live weight)
Farm 1 Farm 2 Farm 3
Production system conventional/ closed system
ecological/ fattening stable
ecological/ fattening stable
Housing system barren enriched enriched Sample size
(in total)
160 106 36
Pigs/pen 19 54 (< 50 kg LW) 18 (> 50 kg LW)
10
Space/pig 0.92m² 0.89m² (< 50 kg LW) 2.67m² (> 50 kg LW)
8.33m²
Pen design inside pens floor: half-planed, half-perforated/ no bedding
inside/outside pens floor: straw bedding
inside/outside pens floor: straw bedding/ soil-based rooting area
Feeding ad libitum pelleted feed
ad libitum mealy feed/hay racks/vegetables and fruits
ad libitum liquid feed/hay racks/vegetables and fruits
Seasonal influence - yes yes
134
The fattening pigs were tested three times during their fattening whereby the first, second and
third points of testing were at the beginning, middle and end of fattening, respectively. Table
2 shows which data were collected at each point of testing. During data collection, no major
changes in management or housing occurred on the three tested farms.
Table 2: Timings of data collection during fattening and sample sizes per farm (LW=live weight)
Start of fattening
(30 kg LW) Middle of fattening
(75 kg LW) End of fattening
(>100 kg LW) Sample size
per farm (n)
Farm
1
Farm
2
Farm
3
Farm
1
Farm
2
Farm
3
Farm
1
Farm
2
Farm
3
Behavioural
tests 160 106 31 160 106 31 160 106 31
Playing
behaviour 138 54 36 - - - 138 54 36
Body
language
signals
138 54 36 - - - 138 54 36
Saliva-
IgA content - - - - - - 25 15 19
Saliva-
protein
composition
- - - - - - 25 15 20
Diameter of
hippocampi - - - - - - 25 15 20
Astroglia
cell
numbers
- - - - - - 25 15 20
Experimental procedures
Behavioural tests
297 fattening pigs of two batches (farm 1: n=160; farm 2: n=106; farm 3: n=31) were tested in
three human approach (HAT) and three novel object tests (NOT), respectively at the
beginning, middle and end of fattening. The pigs were tested alone in their home pen. Both
behavioural tests were never conducted on the same day but always with a one-day time lag in
between. Each pig was given an acclimation period of two minutes followed by a test period
of three minutes that began when the unknown human entered, or the novel object was
brought into, the home pen. Always a female person whom the fattening pigs did not know
from daily routine work represented the unknown human in the HAT. She wore rubber boots
135
and a clean overall and stood motionless in the middle of the home pen during the whole test
period. The utilized novel objects in the NOT were plastic ducks presented to the pigs in three
different sizes (related to the age and the live weight of the fattening pigs). These plastic
ducks indicated a yellow body colour with a red coloured beak. They were held in the middle
of the pen at the height of the pig's head by the use of a rod and a string. During the entire test
period, always the same observing person noted the analysed variables: the approach latency
(AL), the duration of contacts (DC) and the number of contacts (NC). AL was the time in
seconds that each pig needed to approach the unknown human or the novel object until the
snout came into contact with the human or the novel object. DC represented the summed
seconds in which the pigs touched the human or the novel object with their snouts. The entire
number of snout contacts that occurred during the test phase was shown by NC. If the pigs did
not come into contact with the human or the novel object, an AL of 180 seconds was noted.
Both behavioural tests were performed in the home pens of the respective farms. On Farm 1
the entire home pen (3.70 x 4.70m) was used. The amount of space for the fattening pig’s
testing on Farm 2 measured 6.20 x 4.90m with a roofed and unroofed area. On Farm 3, the
roofed outdoor area was utilised (2.40 x 3.80m) for both behavioural tests.
Video data of playing behaviour and body language signals
The same video data were used for the analysis of playing behaviour and body language
signals. 228 fattening pigs (farm 1: n=138; farm 2, n=54; farm 3, n=36) were videotaped for
two days at the beginning and two days at the end of fattening. The camera systems HeiTel
Digital Video GmbH, Kiel, Germany and AXIS M30-VE Network Cameras were used to
record the pigs’ behaviour. The cameras were positioned above the pens to obtain a complete
overview. All pigs were marked individually with a sign on their backs.
Training of the video observers
Due to the great extent of the video data, it was assessed by four observers. At the beginning
of the video analysis, these observers were intensively trained using video test sequences and
through repeated courses of instruction between the individual evaluation sections. Moreover,
these observers were applied cross-classified and no observer was replaced during the whole
video analysis process.
136
Playing behaviour
The playing behaviour was examined for two days at the beginning and two days at the end of
fattening by an ethogram (Table 3) basing on previous literature research. The continuous
sampling method was used to note the duration (s/h) of each playing behaviour sequence. If
several pigs showed playing behaviour simultaneously, the video sequence was analysed for
one pig and then rewound and analysed for the other pigs. Relating to the testing environment,
the collected video material was analysed following the different play categories (Table 3) for
eight hours per day for farm 1 and accordant to the season for the daylight hours for farms 2
and 3. In order to obtain a comparable attribute, the total duration of playing behaviour was
divided by the light day per hour, resulting in the total duration of playing behaviour in
seconds per hour. This procedure was also conducted to obtain the total durations of playing
behaviour (s/h) for each play category (Table 3). The total duration of the locomotor play
(s/h) involved all play categories that were shown exclusively by body movements and the
total duration of the social play (s/h) included all play categories with additional social issues
(Table 3).
137
Table 3: Ethogram of pigs’ playing behaviour
Play category Code Description References
Locomotor play Pivot a Gambolling or twirling the body
by 1 to 360° Brown et al. (2015), Chaloupkova et al. (2007), Donaldson et al. (2002), Newberry et al. (1988)
Hop b Jumping off with the hind legs and angling the forelegs
Welker (1961)
Scamper c Energetic running (from 1m or along the whole pen)
Brown et al. (2015)
Flop d Dropping from an upright to a lying or sitting position on the pen floor
Brown et al. (2015), Chaloupkova et al. (2007), Donaldson et al. (2002)
Roll e Rolling lengthwise over the back Defined for this study Head-shaking j Moving the head back and forth
quickly with ears flapping Reimert et al. (2013)
Social play Non-harmful fighting g Heads and mouths playfully
touching each other (no pushing anti-parallel/ without apparent aggression)
Brown et al. (2015)
Chase h Chasing other pigs playfully (without apparent aggression)
Welker (1961)
Invitation i Pushing the snout with minimal or moderate force into another pig’s body (if playing resulted afterwards)
Martin et al. (2015)
Body language signals
The fattening pigs’ body language signals were examined for two days at the beginning and
two days at the end of fattening followed by an ethogram (Table 4) based on previous
literature research. Through the usage of the scan-sampling method (from the beginning of the
daylight once an hour during the same) the tail respectively ear posture of each fattening pig
was recorded whether they were recognisable, and whether the pigs were walking or standing.
Tail postures were determined as either “curled-up” (the tail forms a loop above the back of
the pig), “hanging” (the tail is neither curled up nor raised but hanging down), “raised” (the
tail is raised but not curled), “tucked-under” (the tail is between the hind legs) or “wagging”
(the tail is wagging). Ear postures were classified as “ears directed forwards” (ears are
directed forwards), “ears directed backwards” (ears are directed backwards), “ears mixed”
138
(one ear is directed forward and one ear backward) or “ears directed laterally” (ears are
directed to the side, neither forwards nor backwards) (Table 4).
Table 4: Ethogram of the different tail- and ear postures and presumed related indicated affective state
Body part Tail/ear
posture
Description Presumed
indicated
affective state
Reference
Tail
Curled-up
Tail forms a loop above the back of the pig
Positive e.g. McGlone et al. (1990)
Hanging
Tail is neither curled up nor raised but hanging down
Negative/Neutral Paoli et al. (2016); Guthrie (1971)
Raised Tail is raised but not curled
Positive/Negative Scheurmann (1974); Reefmann (2009)
Tucked-under Tail is between the hind legs
Negative Groffen (2012)
Wagging Tail is wagging Positive/Negative Kleinbeck and McGlone (1993); Groffen (2012)
Ear
Forwards Ears are directed forwards
Positive/Negative Reefman et al., 2009; Raoult and Gygax (2018)
Backwards Ears are directed backwards
Positive/Negative
Windschnurer et al., 2009; Reimert et al. (2012)
Mixed
One ear is directed forward and one ear backward
Negative Reefmann et al. (2009)
Laterally
Ears are directed to the side, neither forwards nor backwards
Positive/Negative Reefmann et al. (2009);
Physiological data of the present thesis
Out of the 302 fattening pigs examined in total, 60 animals were selected (farm 1: n=25; farm
2: n=15; farm 3: n=20) which definitely provided hippocampal data and less than 15 %
missing variable values (Hair et al., 2014) of all other behavioural (behavioural tests, playing
behaviour and body language signals) and physiological (salivary immunoglobulin A-content
139
respectively protein composition) parameters. This dataset was utilised to analyse the most
appropriate behavioural or physiological parameter to estimate the latent variable of the pigs’
positive affective state by using the partial least squares structural equation modelling method.
Saliva samples
Two saliva samples per pig were collected at the end of fattening when the pigs had a body
weight of about 100 kg. A synthetic fibre role was used for each sample (Cortisol-Salivette®,
Sarstedt AG & Co, Nümbrecht – Germany) and was presented to the pigs with a tweezer. The
fattening pigs had to chew it until it was completely salivated. Then, the samples were marked
individually and frozen immediately.
Saliva - Immunoglobulin A (IgA)
For the analyses of IgA, one of the saliva samples of each pig was thawed at room
temperature and centrifuged at 1000xg for 2 minutes. Hereafter, the supernatant was analysed
with a direct quantitative sandwich-ELISA-Kit for pig-IgA (Celltrend GmbH, Luckenwalde,
Germany).
Saliva - Protein composition
After thawing the insalivated synthetic fibre role at room temperature and centrifuging at
1000xg for 2 minutes, the general protein content (GPC) (µg/ml) were analysed by
colorimetric detection by means of a bicinchoninic acid-based protein assay kit (Pierce™
BCA Protein Assay Kit, ThermoFisher Scientific Inc., Waltham, USA) and the total number
of bands (TNB) and total band intensity (TBI) (px) through a one-dimensional, SDS-
polyacrylamide gel electrophoresis, followed by a modified method of Lamy et al. (2008).
Eight to 16% Mini-PROTEAN® TGX™ Precast Protein Gels (Bio-Rad Laboratories, Inc.,
Hercules, USA) were used to analyse 20 µg of protein per pig in a dual approach. The
required values including this protein amount were identified through the further results of the
general protein content. The samples were concentrated with a concentrator (Concentrator
plus, Eppendorf AG, Hamburg, Germany) if a greater saliva volume was needed to reach 20
µg protein as the wells had a maximal capacity of 20 µl. After implementation on the gels, the
samples ran at 40 volts for 15 minutes and then at 100 volts for 75 minutes. The gels were
then fixed with 50 ml of fixing solution for 20 minutes and stained with 50 ml of staining
140
solution for 30 minutes. Subsequently, the gels were photographed after being destained in
10% acetic acid for 46 hours.
The GelAnalyzer 2010a software (Lazarsoftware) was used to analyse the pictures by
extracting two parameters: the TNB for each lane and each pig and the intensity of all bands
TBI for each pig. The bands weights were specified by a marker (Precision Plus Protein™
Dual Xtra Standard, Bio Rad Laboratories GmbH, München, Germany) and then the intensity
of each individual band was identified by the area under the peak (AUP) in pixels (px). The
TNB depicts the mean of the sum of bands of both lanes for each pig, whereas the TBI
represents the mean of the intensities of all bands of both lanes of each pig.
Slaughter organs - hippocampi
Both hemispheres of the brains were taken separately from the carcasses during the
slaughtering process and were fixed in 4% paraformaldehyde for three days. After fixation,
the hippocampi of each hemisphere were removed, and 5 mm vertical slices were cut out and
transferred to 1% paraformaldehyde. Afterwards, the samples were embedded in paraffin
blocks and cut again into 50 µm slices by use of a microtome (RM 2155, Leica Biosystems,
Nussloch, Germany) and stained with a glial fibrillary acidic protein (GFAP)
immunohistochemistry. Then, four sections of every individual (two sections per hemisphere)
were analysed and averaged. The diameter from a line between the ventral expression of the
dentate gyrus and the CA3 region of the hippocampus proper to the middle of the CA1 region
was measured for the analyses of the hippocampal size. Pictures of the dorsal expression of
the dentate gyrus were taken at 10-fold magnification (Axiophot, Carl Zeiss Microscopy
GmbH, Jena, Germany) by the use of the AxioVision Rel. 4.8 software (Carl Zeiss
Microscopy GmbH, Jena, Germany) to detect the astroglia cell number. Hereafter, the
staining intensity of the pictures was assessed semi-quantitively with the ImageJ software
(National Institutes of Health, Bethesda, Maryland, USA). The resulting GFAP-staining-
pixel intensities were negatively correlated with the number of GFAP-expressing cells.
Statistical procedures
Statistical analyses for the behavioural tests, playing behaviour and body language signals
were performed with the software package SAS® 9.4 (SAS Institute Inc., 2017) using linear
mixed models respectively generalised linear models. The fit statistics AICC “Akaike’s
information criterion corrected” and the BIC “Bayesian information criterion” were proved to
141
evaluate the fitting of the statistical models. Fixed effects were added in a stepwise manner to
the models. The models with the smallest AICC and BIC values were chosen. The
significance of differences in the least square means was adjusted with the Bonferroni-
correction. Statistical significance was determined at p<0.05.
The SmartPLS 3.0 software (Ringle et al., 2015) was used to generate and calculate the partial
least squares structural equation model to investigate the influence of the fattening pigs’
positive affective state on the variety of behavioural and physiological parameters and to
analyse the most appropriate parameters to estimate the latent variable of the pigs’ positive
affective state. The evaluation was based on a variety of recommended quality criteria by Hair
et al. (2017).
Behavioural tests
The behavioural test data were not normally distributed. Thus, all data were log10 (X + 1)
transformed to obtain normality of residuals of the used linear mixed model (PROC MIXED).
The model used included the fixed effects farm (1-3), batch (1, 2) nested in farm, points of
testing (beginning, middle and end of fattening) nested in farm and gender (female, male)
together with a random effect of each individual pig nested in farm, batch and gender. The
residuals of the linear mixed models of all behavioural variables of both behavioural tests
were correlated through the Pearson correlation coefficient (PROC CORR).
Playing behaviour
The playing behaviour data showed no normal distribution. All data were log10 (X + 1)
transformed to obtain normality of residuals of the linear mixed models used (PROC
MIXED). The chosen model included the fixed effects observer (1-4), farm (farm 1, farm 2
and farm 3), gender (female, male), points of testing (first fattening phase, final fattening)
nested in farm, day (1, 2) nested in farm together with a random effect of each individual pig
nested in farm and gender and was applied to analyse the data of the total duration of playing
behaviour (s/h) and total duration of locomotor respectively social play (s/h). The same model
with the additional fixed effect location nested in farm was used to compare the total duration
of playing behaviour (s/h) between the inside and outside area of farms 2 and 3; for this, a
dataset was used that included only these two farms. As there were too few observations in
some play categories to apply statistical models, the differences between the farms within the
individual play categories were analysed descriptively using the Kruskal-Wallis test.
142
Body language signals
The tail respectively ear postures were noted as binomial data (0 = respective tail or ear
posture was not shown; 1 = respective tail or ear posture was shown) and were analysed using
the GLIMMIX procedure with a binomial distribution (link-function = logit). Fixed effects
were added to the model in a stepwise manner. The model for the data of the curled-up tails,
hanging tails and ears directed forwards included the fixed effects farm (1-3), points of testing
(beginning respectively end of fattening) nested in farm, gender (female, male) and a random
effect of each individual pig nested in farm. The model for the data of the raised tails,
wagging tails and laterally directed ears included the fixed effects of farm (1-3), gender
(female, male) and a random effect of each individual pig nested in farm. Statistical
significance was determined at p<0.05 and the significance of differences in the least square
means was adjusted with the Bonferroni-correction. Due to rare occurrences, statistical
models could not be used for the data of the tucked-under tails, ears directed backwards and
ears mixed and therefore these results are not shown.
Influence of the pigs’ positive affective state on behavioural and physiological parameters
The partial least squares structural equation modelling method (PLS-SEM) was applied to
analyse the most appropriate parameters to estimate the latent variable of the fattening pigs’
primarily positive affective state and to evaluate the latent structures in between the
behavioural and physiological parameters.
General scheme of a structural equation model (SEM)
A SEM includes two types of latent variables which indicate the relationships between them.
There are not directly measurable latent variables and latent measurement variables, which
enable an estimation of the former latent variables. A structural model which consists of
endogenous respectively exogenous latent variables and at least two measurement models
(latent measurement variables and their self-characterising indicators) depict a SEM. The
relationships (path coefficients) between the endogenous respectively exogenous latent
variables are represented by connected arrows to each other. Endogenous (dependant) latent
variables can be affected by other latent variables, irrelevant whether they are endogenous or
exogenous, though exogenous (independent) latent variables cannot be influenced by other
latent variables. Accordingly, the coefficient of determination (R²), which indicates to which
143
extent the latent variable is explained by other latent variables, is exclusively analysed for
endogenous latent variables.
Hierarchical Component Models (HCM)
The SEM of the present thesis needed an operationalisation at a higher abstraction level so a
hierarchical component model (HCM) with two levels of abstraction was conducted to
analyse the influence of the pigs’ positive affective state on different behavioural and
physiological parameters. The higher-order component (HOC) which depicts the more
abstract level and the lower-order components (LOC) which include the sub-dimensions of
the higher component form the two levels of abstraction of a HCM. In the present thesis, the
pig’s affective state represented the HOC and the different behavioural and physiological
parameters illustrated the LOC. A reflective-reflective HCM was used which included a
reflective relationship between the HOC and LOC whereby all constructs of LOC were
indicated by reflective measurement models. A reflective measurement model implies that the
indicators represent the underlying construct and that the causality of the construct is directed
to the indicators. In order to form the HOC measurement model, all indicators of the LOC
were used repeatedly to the HOC (Hair et al., 2017).
Assessment of the SEM
The evaluation involves a two-stage process whereby firstly the measurement models are
proven, followed by the evaluation of the structural model. Quality criteria of the
measurement models involve the internal consistency reliability (composite reliability) and
convergent validity (indicator reliability and average variance extracted (AVE)). Regarding
the consistency reliability, the composite reliability takes different loadings of indicators into
account and illustrates values between 0 and 1. In exploratory research studies, composite
reliability should assume values between 0.6 and 0.7. The indicator reliability evaluates how
sufficient a latent variable is estimated by an indicator with regard to the convergent validity.
AVE depicts the extent to which a latent construct explains the variance of its indicators.
AVE should achieve values above or equal to 0.5.
Important quality criteria of the structural model include the examination of the coefficients of
determination (R²) and the relevance and statistical significance of path coefficients. The path
coefficients characterise the relationships between the constructs in the structural model and
comply with the standardised regression coefficients in the regression analysis. Path
144
coefficients verify or refute the pre-hypothesised relationships between the constructs in the
model by showing values between -1 and 1. For this, the sign and magnitude of the path
coefficients have to be considered (Henseler et al., 2009). An estimated path coefficient near
to +1 indicates a strong, positive relationship between examined constructs (and vice versa for
negative values), which is typically statistically significant. The weaker the relationship, the
closer the estimated path coefficient is to 0. Very low values close to 0 are generally not
statistically significant (Hair et al., 2017) whereby the statistical significance is proven by the
bootstrapping method. The coefficients of determination (R²) illustrate the proportion of the
variance of an endogenous construct that is explained by all precursor constructs
corresponding to the endogenous construct. The construct is explained all the better, the
higher the R²-values are (Hair et al., 2017). R2-values of 0.25, 0.50 and 0.75 are assessed as
weak, moderate and substantial, respectively (Henseler et al., 2009; Hair et al., 2011).
145
REFERENCES
Brown, S. M., Klaffenböck, M., Nevison, I.M., Lawrence, A. B., 2015. Evidence for litter
differences in play behaviour in pre-weaned pigs. Applied Animal Behaviour Science
172, 17–25.
Chaloupková, H., Illmann, G., BartoŠ, L., Špinka, M., 2007. The effect of pre-weaning
housing on the play and agonistic behaviour of domestic pigs. Applied Animal
Behaviour Science 103, 25-34.
Donaldson, T. M., Newberry, R. C., Špinka, M., Cloutier, S., 2002. Effects of early play
experience on play behaviour of piglets after weaning. Applied Animal Behaviour
Science 79 (3), 221–231.
Groffen, J., 2012. Tail posture and motion as a possible indicator of emotional state in pigs.
Animal Science program, Exam work in Animal Science EX 0567, SLU and
Wageningen University (WUR).
Guthrie, R. D., 1971. A new theory of mammalian rump patch evolution. Behaviour 38, 132-
145.
Hair, J.F., Hult, G.T.M., Ringle, C.M., Sarstedt, M., Richter, N.F., Hauff, S., 2017. Partial
Least Squares Strukturgleichungsmodellierung (PLS-SEM). Verlag Franz Vahlen
GmbH München.
Hair, J.F., Ringle, C.M., Sarstedt, M., 2011. PLS-SEM: Indeed a silver bullet, Journal of
Marketing Theory and Practice, 19, 139-151.
Hair, J.F., Tomas, G., Hult, M., Ringle, C.M., Sarstedt, M., 2014. A primer on partial least
square structural equation modeling (PLS-SEM). SAGE Publications, Inc. Thousand
Oaks, California, USA.
Henseler J., Ringle, C.M., Sinkovics, R.R., 2009. The use of partial least squares path
modelling in international marketing. Advances in International Marketing. 277–320.
Kleinbeck, S., McGlone, J.J., 1993. Pig tail posture: A measure of stress. Technical
University Agricultural Science Technical Report T-5-327: 47–48.
146
Lamy, E., da Costa, G., Capela e Silva, F., Potes, J., Coelho, A.V., Baptista, E.S., 2008.
Comparison of electrophoretic protein profiles from sheep and goat parotid saliva.
Journal of chemical ecology 34, 388–397.
Martin, J. E., Ison, S. H.; Baxter, E. M., 2015. The influence of neonatal environment on
piglet play behaviour and post-weaning social and cognitive development. Applied
AnimalBehaviour Science 163, 69–79.
McGlone, J.J., Sells, J., Harris, S., Hurst, R.J., 1990. Cannibalism in growing pigs: effects of
tail docking and housing system on behaviour, performance and immune function.
Journal of Agricultural Science and Technology. Report No. T-5-283. Texas Tech
University, 69–71.
Newberry, R. C., Wood-Gush, D.G.M., Hall, J. W., 1988. Playful behaviour of piglets.
Behavioural Processes 17 (3), 205–216.
Paoli, M. A.; Lahrmann, H. P.; Jensen, T.; D'Eath, R. B., 2016. Behavioural differences
between weaner pigs with intact and docked tails. In: anim welf 25 (2), 287–296.
Raoult, C.M.C., Gygax, L., 2018. Valence and Intensity of Video Stimuli of Dogs and
Conspecifics in Sheep: Approach-Avoidance, Operant Response, and Attention.
Animals 8(7), 121.
Reefmann, N., Bütikofer Kaszàs, F., Wechsler, B.; Gygax, L., 2009. Ear and tail postures as
indicators of emotional valence in sheep. In: Applied Animal Behaviour Science 118
(3-4), 199–207.
Reimert, I., Bolhuis, J., Kemp, B., Rodenburg, T., 2012. Indicators of positive and negative
emotions and emotional contagion in pigs. Physiology Behaviour 109, 42–50.
Reimert, I., Bolhuis, J. E., Kemp, B., Rodenburg, T. B., 2013. Indicators of positive and
negative emotions and emotional contagion in pigs. Physiology and Behavior, 109,
42–50.
Ringle, C.M., Wende, S., Becker, J.M., 2015. SmartPLS 3 (Software).
SAS 2017. SAS Institute Inc. Cary, NC, USA.
147
Scheurmann, E., 1974. Ursachen und Verhütung des gegenseitigen Besaugens bei Kälbern.
Tierärztliche Praxis. 2, 389–394.
Welker, W.I., 1961. An analysis of exploratory and play behavior in animals. In Fiske, D. W.,
and Maddi, S. R. (eds.) Functions of Varied Experience, Dorsey, Homewood, Ill.
Windschnurer, I., Barth, K., Waiblinger, S., 2009. Can stroking during milking decrease
avoidance distances of cows towards humans? Animal Welfare 18, 507–513.
DANKSAGUNG
Mein besonderer Dank gilt meinem Doktorvater Herrn Prof. Dr. Joachim Krieter für die
Überlassung des interessanten Promotionsthemas, die gute wissenschaftliche Betreuung und
die Möglichkeiten meine Ergebnisse auf nationalen sowie internationalen Konferenzen
präsentieren zu können.
Bei Frau Prof. Dr. Nicole Kemper möchte ich mich herzlich für die Übernahme des
Koreferates bedanken.
Ein besonderer Dank für die Entwicklung des wunderbaren Projektes „FeelGood“ gilt Frau
Dr. Irena Czycholl. Zudem möchte ich ihr und Frau Dr. Kathrin Büttner für die motivierende
Betreuung und für das stets schnelle und sorgfältige Korrekturlesen danken.
Den Betriebsleitern der drei Betriebe danke ich für die Bereitstellung ihrer Mastschweine und
das allseits gewährte Vertrauen während der gesamten Datenaufnahme.
Für die schöne Zeit am Institut, dem einen oder anderem „Bauernball“- oder „Born for Korn“-
Besuch, gemeinsamen Grillen, unterhaltsamen Schlachthofbesuchen und gegenseitiger
Unterstützung möchte ich allen derzeitigen und insbesondere ehemaligen Kollegen danken!
Ein großes Dankeschön für eine wunderbare Zeit, jederzeit sich sehr ergänzende
Zusammenarbeit und das stets offene Ohr gilt dem anderen „FeelGoodie“ Farina, eine bessere
Partnerin hätte ich mir für diese enge Teamarbeit nicht wünschen können.
Auch meinen lieben Freunden außerhalb des Institutes möchte ich für die ständige Motivation
und Unterstützung während der gesamten Promotionszeit danken.
Ein spezieller Dank gilt meinem Freund, der mich mit seinem unermüdlichen Drang und der
Motivation „Im Auftrag der Landwirtschaft“ zu arbeiten, unterstützte und stets zu motivieren
wusste.
Mein größter Dank gilt meiner Familie. Ihr habt mich immer unterstützt, an mich geglaubt
und mir in jeglicher Form Rückhalt geboten. Ohne euch wäre ich nicht dorthin gekommen,
wo ich jetzt bin.
LEBENSLAUF
Persönliche Daten
Name Katja Lisabeth Krugmann
Geburtsdatum 05. März 1991
Geburtsort Lübeck
Beruflicher Werdegang
seit 07/2016 Wissenschaftliche Mitarbeiterin am Institut für Tierzucht und Tierhaltung der Christian-Albrechts-Universität zu Kiel in der Arbeitsgruppe von Herrn Prof. Dr. Joachim Krieter
06/2016 Master of Science Agrarwissenschaften, Christian-Albrechts-Universität zu Kiel
12/2014 Bachelor of Science Agrarwissenschaften, Christian-Albrechts-Universität zu Kiel
Ausbildung
10/2010 – 06/2016 Studium der Agrarwissenschaften an der Christian-Albrechts-
Universität zu Kiel Abschluss: Master of Science Agrarwissenschaften
09/2001 – 07/2010 Leibniz Gymnasium Bad Schwartau Abschluss: Abitur