expert consultation on weighting factors of criteria for assessing environmental enrichment...
TRANSCRIPT
Expert consultation on weighting factors of
criteria for assessing environmental enrichment
materials for pigs
Marc B.M. Bracke *, Johan J. Zonderland, Edwin J.B. Bleumer
Animal Sciences Group of Wageningen, Wageningen University and Research Centre,
PO Box 65, 8200 AB Lelystad, The Netherlands
Accepted 8 May 2006
Available online 16 June 2006
Abstract
EC directive 2001/93/EC states that pigs must have permanent access to a sufficient quantity of material
to enable proper investigation and manipulation activities. This directive requires further interpretation. In
order to facilitate the further implementation of the directive into national, Dutch legislation a preliminary
model was constructed to assess the value of different enrichment materials for pigs. Using an e-mail
questionnaire expert opinion was elicited in order to ‘validate’ the preliminary model. In total 8 senior pig
welfare experts assessed 33 assessment criteria ordered according to the conceptual framework underlying
the preliminary model. Kendall’s coefficient of concordance of the experts without missing values (n = 6)
was 0.41 (P < 0.001), which is moderate. Assessment criteria that generated the highest weighting factor
scores included exploration, animal–material interactions (AMI), tail and ear biting, and rooting (median
expert scores higher than 8.0). The Spearman’s rank correlation between the expert median scores and the
preliminary model was 0.63 (P < 0.001).The scope for modelling and implications for ethical and political
decision-making are discussed.
# 2006 Elsevier B.V. All rights reserved.
Keywords: Enrichment; Pigs; Housing; Toys; Validation; Assessment criteria
www.elsevier.com/locate/applanim
Applied Animal Behaviour Science 104 (2007) 14–23
DOI of related article: 10.1016/j.applanim.2006.05.005.
* Corresponding author. Tel.: +31 320 238205; fax: +31 320 238094.
E-mail address: [email protected] (Marc B.M. Bracke).
0168-1591/$ – see front matter # 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.applanim.2006.05.006
1. Introduction
Environmental enrichment is important for animal welfare, as farm animals are often kept in
barren environments (e.g. Young, 2003). For example, the Scientific Veterinary Committee of the
European Commission (SVC, 1997) recommended providing the pigs with materials for
investigation and manipulation, which may be bedding material or earth floors suitable for rooting.
In 2001 the European Commission adopted a directive (2001/93/EC) which states that: ‘‘Pigs
must have permanent access to a sufficient quantity of material to enable proper investigation and
manipulation activities, such as straw, hay, wood, sawdust, mushroom compost, peat or a mixture
of such, which does not compromise the health of the animals’’. As of January 2005 this
requirement applies to all holdings. This directive has been implemented in most EU
memberstate’s legislation, including the Dutch legislation.
A problem with the directive is that it leaves toomuch room for interpretation. It isnot sufficiently
clear what is proper investigation and manipulation, and what is required in this regard for providing
acceptable enrichment. The directive specifies a number of materials as examples. However, the
value of enrichment material is most likely determined not only by the type of material, but also by
other properties of the material such as the amount and frequency of material provision.
Memberstates and individual farmers may, therefore, differ in their interpretation of the
directive. This may compromise the level playing field, i.e. fair economic conditions for farmers,
across EU memberstates.
When specifying the requirements implied by the directive the Dutch Ministry of Agriculture,
Nature and Food Quality would prefer to formulate these in terms of goals rather than in terms of
means. Ideally, this means that measurements should be taken on the farm to establish the welfare
impact of the enrichment materials, e.g. the duration of object play and/or the degree of reduction
in tail biting. Since this was not considered feasible at this point in time, the Ministry decided to
nevertheless avoid a pure prescription of means (e.g. in the form of a limited list of allowed
enrichment materials) and opted for a kind of prescription of intermediate goals, in the form of
assessment criteria that were linked to enrichment materials on the one hand and known welfare
performance measures on the other hand within the context of a model to assess the enrichment
value of enrichment materials for pigs (see Bracke et al., 2006). This could allow farmers to meet
stated objectives through different routes in farm specific solutions and would, furthermore,
allow stimulating innovations towards a better provision of environmental enrichment for pigs.
The model, called RICHPIG, allows calculating scores for the (relative) value of different
enrichment materials using a list of assessment criteria and procedures described earlier for
welfare assessment in pregnant sows (Bracke et al., 2002a) and tail biting (Bracke et al., 2004).
For the work presented in this paper a preliminary version of RICHPIG was used to identify
assessment criteria, based on a systematic analysis of (part of) the scientific literature. RICHPIG
allowed the calculation of weighting factors indicating the relative importance of the assessment
criteria based on the analysis of the scientific information in the database. The model’s weighting
factors were compared with the scores given by the experts. This was done in order to generate a
second validation of the model (the first validation being a comparison with an expert evaluation
of enrichment materials, Bracke et al., 2007). It serves as a statement of how internationally
recognised experts in the field assess the different assessment criteria for evaluating
environmental enrichment in pigs, given the present state of science.
The aim of this paper/questionnaire, therefore, was to examine expert opinion concerning the
weighting of assessment criteria for assessing enrichment materials for pigs and to determine
how the scores relate to the scores of the preliminary RICHPIG model.
M.B.M. Bracke et al. / Applied Animal Behaviour Science 104 (2007) 14–23 15
2. Materials and methods
2.1. The questionnaire
In December 2004, a written email questionnaire was sent to nine pig-welfare experts from six different
EU countries. The experts had been selected based on their expertise and (presumed) influence on policy
making in their home country, and had agreed to participate in the previous questionnaire on assessing
enrichment materials.
One expert did not return the questionnaire. In total eight experts responded from five different EU
countries: four from the Netherlands, one from Germany, one from Denmark, one from Sweden and one
from the United Kingdom.
Thirty-three assessment criteria were presented to the experts. These criteria were grouped in the
following classes (where n indicates the number of criteria in the class):
a. Object design criteria (‘causes’), such as smell, taste and accessibility (n = 7);
b. Behavioural elements of positive animal-material interactions (AMI), such as rooting, nosing and biting
(n = 8);
c. Biologically functional objectives of AMI, namely exploration/learning and foraging (n = 2);
d. Amount of manipulation of enrichment materials (EMats) and other items, including (overall) AMI, pen-
directed behaviour and tail biting (n = 3);
e. Other (positive and negative) consequences, such as effects on production, stress and fear (n = 8);
f. Object performance consequences, such as that the material is changed (e.g. destructed), makes a sound
and does not get soiled (n = 5).
This classification was based on the underlying conceptual framework for (welfare) assessment (Anon-
ymous, 2001). The assessment criteria were presented in a fixed order and sorted according to the specified
classification.
The experts were asked to give a score for the relative importance of each criterion on a scale between 0
(not at all important) and 10 (very important), where each weighting factor score expressed the degree to
which the criterion was considered relevant for assessing the (duration, incidence and intensity of)overall
enrichment value of a wide range of different enrichment materials for pigs. The experts were also asked to
specify any criteria they felt were missing and to specify their considerations for assigning weighting
factors.
2.2. WAFL-05
In addition, a highly similar list of 30 assessment criteria was presented at the 3rd International
Workshop on the Assessment of Animal Welfare at Farm and Group Level (22–24 September 2005,
Vienna, Austria; Bracke, 2006). Participants in the working group meeting on semantic modelling, which
were mainly European scientists working on welfare assessment in farm animals, were requested to give
a score on a scale from 0 (worst) to 10 (best) to the following assessment criteria: smell, taste, visibility,
novelty, nutritious, accessible, nose, root, bite, pull, chew, explore/learn, forage, animal–material
interactions, pen-directed behaviour, abnormal behaviour, tail biting, (other) penmate-directed beha-
viour, activity, production, fear, stress, aggression, disturbance, health, sound, move, change, ingest,
hygiene.
2.3. Statistics
Pearson correlation coefficients (R) and Spearman rank correlation coefficients (Rho) were used to
compare the various sets of scores. R2 is given as indicator of the percentage of variance explained by the
(linear) correlation model.
M.B.M. Bracke et al. / Applied Animal Behaviour Science 104 (2007) 14–2316
Residuals of the Pearson’s correlation coefficients were used to identify individual weighting factors
where the experts as a group deviated the most from the model scores.
Sign tests were used to determine for which assessment criteria the model differed significantly from the
experts in assigning weighting factors. To this end the number of expert scores above and below the model
score was counted per treatment.
Kendall’s coefficient of concordance (W) was used to examine the degree of consensus between the
experts, and between the experts and the preliminary model.
The Friedman two-way analysis of variance by ranks and the post hoc multiple comparisons test
(ANOVA over ranks in Genstat) were used to determine whether there were significant differences in ranks
between weighting factors of assessment criteria (Siegel and Castellan, 1988, pp. 174–183). Since the
Friedman test cannot handle missing values, we used overall medians to represent expert opinion. The data
were analysed using Genstat (Anonymous, 1993).
3. Results
3.1. Weighting factors of assessment criteria
Different types of arguments were given to support weighting factor scores. One major type of
argument referred to what is attractive to the pig in the sense that it increases animal–material
interactions (AMI). Another type of argument referred to the nature of the animal (e.g. that smell
is very important for pigs, that pigs spend much time chewing, and that there is an age effect in
that young animals play more). Some experts pointed out that scores depend on the
(interpretation of the) definitions (e.g. ‘rooting’, ‘health’ and ‘production’) and possible
interactions (e.g. the importance of ‘health’ may depend on the level of ‘stress’).
Table 1 shows the median expert weighting factor scores for the different assessment criteria.
The Friedman analysis was significant (P < 0.05) indicating that there were significant
differences between the criteria according to the experts. The results of post hoc comparisons are
also shown in Table 1 as ‘significance levels’.
Kendall’s coefficient of concordance (W) of the experts without missing values (n = 6) was
0.41 (P < 0.001). For the criteria that were included in the preliminary model, W of the 6 experts
was 0.36 (P < 0.001). Including the preliminary model scores as if it was an expert increased W
to 0.38 (P < 0.001).
Assessment criteria that generated the lowest weighting factor scores included production,
carrying, visually inspecting, sound producing (median expert scores smaller than 4.5).
Assessment criteria that generated the highest weighting factor scores included exploration,
AMI, tail and ear biting, and rooting (median expert scores higher than 8.0).
3.2. Correlation between expert opinion and model scores
The Pearson correlation between the preliminary model scores and the median expert scores
for the weighting factors of the different assessment criteria is 0.64 (R2 = 0.41). Rho is 0.63
(P < 0.001).
Criteria with the largest residuals included: production, visually appealing, sound produ-
cing, nutritiousness, chewing, where the experts scored lower than the model (residuals < �1.0),
and nosing, stress, aggression, health and penmate-directed behaviour, where the experts
scored higher than the model (residuals > 1.0). The assessment criterion ‘production’
(expert median = 1.5; preliminary model score = 1.7, see Fig. 1) had a notable residual score
of �3.19.
M.B.M. Bracke et al. / Applied Animal Behaviour Science 104 (2007) 14–23 17
M.B
.M.
Bra
ckeet
al./A
pp
liedA
nim
al
Beh
avio
ur
Scien
ce1
04
(20
07
)1
4–
23
18
Table 1
Assessment criteria listed according to the conceptual assessment framework underlying the model
Brief description of the criterion Assessment criterion Expert
score
Model
score
Residuals Significance
level
Object design criteria (‘causes’)
The degree to which the enrichment material has an attractive smell
for pigs.
Smelling, odorous 7.0 5.30 . . .. . .fghij
The degree to which the material is palatable. Palatability, flavour, taste 6.5 3.30 .. . .efghij
The degree to which the material is visually stimulating for the pigs
(i.e. eliciting visual orientation and investigation).
Visually appealing 3.5 1.50 �1.67 ab. . .. . ..
The degree to which the material is novel (e.g. because it is renewed). Novel/renewed 8.0 6.95 . . .. . .hij
The degree to which the material is functional for performing different
behavioural elements and objectives (e.g. explore, learn, forage).
Multifunctional 7.0 * ..cdefgh..
The degree to which the material is nutritious Nutritiousness 5.0 3.90 �1.13 abc. . .. . ..
The degree to which the material is (simultaneously) accessible to
the pigs
Accessibility 7.5 5.50 . . ...fghij
Behavioural elements of positive animal–material interactions (AMI)
The degree to which the material is suitable for nosing Nosing 8.0 2.50 1.39 . . .. . .ghij
The degree to which the material is suitable for pushing Pushing 6.5 * . . ..efghi.
The degree to which the material is suitable for rooting Rooting 8.5 11.85 . . .. . ...ijThe degree to which the material is suitable for biting (with the
front teeth)
Biting 8.0 6.55 . . ...fghij
The degree to which the material is suitable for pulling Pulling 5.5 3.10 abcde. . ...The degree to which the material is suitable for shaking Shaking 4.5 * abcd. . .. . .
The degree to which the material can be carried around Carrying 3.0 * a. . .. . .. . .
The degree to which the material can be chewed (with the molars) Chewing 6.5 8.05 �1.04 . . ...fghij
Biologically functional objectives of AMI
The degree to which the material is functional for exploration and
learning (incl. information gathering, learning motor skills, solving
problems)
Explore/learn 9.5 12.20 . . .. . .. . .j
The degree to which the material is functional for the expression
of foraging behaviour leading to food consumption
Foraging 7.5 7.50 . . .. . ..hij
Amount of manipulation of EMats and other items
The degree to which the pigs interact with the material (thereby
providing occupation)
Animal–material interaction 9.5 12.40 . . .. . .. . .j
M.B
.M.
Bra
ckeet
al./A
pp
liedA
nim
al
Beh
avio
ur
Scien
ce1
04
(20
07
)1
4–
23
19
The degree to which the pigs interact with the pen walls, floor and
pen fittings
Pen-directed behaviour 5.0 2.70 .bcdef. . ..
The degree to which the pigs are engaged in ear biting and/or tail
biting (both behaviourally and in terms of clinical wounds)
Tail and ear biting 9.0 12.45 . . .. . .ghij
Other (positive and negative) consequences
The degree to which the material affects (other, i.e. non AMI, pen-
or penmate-directed) activity levels (including e.g. locomotor play)
Activity 6.5 3.30 . . .defgh..
The degree to which the material affects production levels (but only
in as far as this is indicative of the welfare of the pigs)
Production 1.5 1.70 �3.19 a. . .. . .. . .
The degree to which the material affects fear levels Fear 7.0 1.80 . . .defgh..
The degree to which the material affects stress levels Stress 8.0 2.60 1.37 . . ...fghij
The degree to which the material affects penmate-directed behaviour
(excluding tail- and ear biting)
(Other) harmful social behaviour 8.0 3.70 1.1 . . .. . .ghij
The degree to which the material affects aggression levels Aggression 8.0 3.00 1.27 . . ...fghij
The degree to which the material affects the level of disturbance
of penmates
Disturbance of other pigs 6.0 1.30 .bcdefg. . .
The degree to which the material affects the health status of the pigs Health 8.0 3.30 1.2 . . .. . .ghij
Object performance consequences
The degree to which the material produces (attractive) sounds Sound producing 4.0 1.30 �1.26 abc. . .. . ..The degree to which the material moves (e.g. in response to AMI) Moveability 7.0 1.20 . . ...fghij
The degree to which the material is changed and/or destructed by
the pigs
Changed 8.0 6.30 . . .. . ..hij
The degree to which the material is ingested by the pigs Ingestion 7.0 5.20 . . ..efghi.
The degree to which the material is/gets soiled Hygiene/soiling 7.0 4.60 ..cdefgh..
Scores in the column ‘Expert score’ represent expert medians (n = 8). Significance levels represent results of the post hoc Friedman test over expert scores, where criteria with
different characters differ significantly at P < 0.05. Residuals above 1.0 and below �1.0 are given of the linear regression of the scores of expert medians and preliminary model
scores. Negative residuals indicate criteria where the model overestimated the experts. Positive residuals indicate criteria where the model underestimated the experts.
*Preliminarily model scores for these criteria cannot be given, because these criteria have been removed from the model as they were subsumed under one or more of the other
criteria in the list. EMats: enrichment materials.
M.B.M. Bracke et al. / Applied Animal Behaviour Science 104 (2007) 14–2320
Fig. 3. Scatter plot and fitted line of the Pearson correlation coefficient and R2 of the preliminary model scores and the
WAFL-05 median expert scores.
Fig. 1. Scatter plot and fitted line of the Pearson correlation coefficient and R2 of the median expert scores and the
preliminary model scores.
Fig. 2. Scatter plot and fitted line of the Pearson correlation coefficient and R2 of the median senior expert scores and the
WAFL-05 median expert scores.
3.3. WAFL-05 results
In total 13 respondents handed in scores for the list of 30 assessment criteria at the WAFL-05
meeting in Vienna.
Low correlation coefficients were found between the median scores from this group of experts
and the previous group, and between this group and the model (respectively, R = 0.61 and
R = 0.33, see Figs. 2 and 3).
The experts at WAFL-05 gave considerably higher scores to ‘production’ (median = 4.0) than
did the previous group of senior experts (median = 1.5).
Compared to the model the experts gave relatively high scores for fear, stress, abnormal
behaviour and health, while they gave relatively low scores for rooting and exploration/learning.
4. Discussion
This paper provides potentially valuable information on criteria to assess enrichment value in
pigs. Such information is relatively rare, because scientific studies normally compare different
enrichment materials rather than assessment criteria (see Van de Weerd et al., 2003 for an
exception). Our study focussed on the comparison of expert opinion and the results of the
preliminary model, called RICHPIG. The work involved European experts, because the work was
conducted in the context of implementing European legislation. As a consequence, however, we
cannot exclude a certain ‘European bias’ in the results (see Fraser, 2003). Furthermore, it should
be noted that the criteria were presented to the experts in a fixed order based on the underlying
classification for welfare assessment. As a consequence carry-over effects cannot be fully
excluded. The results, therefore, may best be regarded as developmental and requiring further
upgrading as the knowledge base or its field of application changes.
Comparing expert opinion with the results of the preliminary model showed that important
criteria include exploration, AMI, tail and ear biting, and rooting (median expert scores higher
than 8.0; preliminary model scores > 11.80). We also found a much lower correlation for
assessment criteria in this study than we have found for assessing enrichment materials (see
Bracke et al., 2007; Rho was 0.63 and 0.97, respectively). This resembles a previous finding,
where the correlation for overall welfare of housing systems for pregnant sows was considerably
higher than the corresponding correlation for weighting factors of attributes (0.92 and 0.73,
respectively; Bracke et al., 2002b). This pattern is also reflected in the Kendall’s W, indicating
that experts agree more on overall assessments than on their respective components. A possible
explanation for this finding may be that studies have focussed on the system/material level. This
seems to suggest that as a general rule ambiguity/‘errors’ at the component level don’t
accumulate into the aggregated level, but that they rather get averaged out in the aggregation
process. An implication of this finding for political decision-making is that it may be
considerably easier to reach consensus on overall items (enrichment materials, housing systems)
than on components (attributes, assessment criteria). In the case of housing systems it was
advised to complement present legislation in terms of attributes with prescriptions at the housing-
system level (Bracke et al., 2002b). In the case of enrichment materials it may be advised to
complement the formulation of prescriptions of goals (using assessment criteria) with a policy
statement as to how this relates to the level of enrichment materials.
Specific criteria where the experts differed from the model included production*, visually
appealing, sound producing, nutritiousness, chewing (experts < the model), and nosing, stress*,
aggression*, health* and penmate-directed behaviour* (experts > the model). Five of these
M.B.M. Bracke et al. / Applied Animal Behaviour Science 104 (2007) 14–23 21
(marked with *) fall in the category ‘other (positive and negative) consequences’. This category
includes 8 items (apart from those marked with * above, also activity, fear and disturbance of
other pigs are included in this group). We also found that compared to the model the respondents
at the WAFL-05 workshop gave relatively high scores for fear, stress, abnormal behaviour and
health. Again, these belong to the group of ‘other consequences’, which may be perceived to be
determinants of welfare generally, rather than of enrichment per se. This could have resulted in
ambiguity (as what is important for welfare generally, doesn’t need to be as important for
enrichment materials in particular, and vice versa), leading to the discrepancies found.
The discrepancy was especially considerable in the case of ‘production’ (expert median = 1.5;
preliminary model score = 1.7; residual score = �3.19). It appeared that some experts had given
a high score to ‘production’, while (most) other experts had given a low score. Verifying the
‘considerations’ given to the scores showed that high scores were related to the view that high
stress levels must have been involved when production levels were reduced. A subsequent e-mail
discussion with the experts involved led to the following hypothesis. Some authors may have
interpreted ‘production’ in relation to enrichment as an indicator of stress, which, when
substantially reduced would indicate considerable stress, resulting in a high weighting factor.
Other experts may have been led by the (current) view that enrichment has never been shown to
have any substantial impact on production, resulting in a low weighting factor. This latter view is
consistent with the procedures underlying the model, as it required scientific findings collected in
the database to generate (increased) weighting factors. Careful (technical) specification of
assessment criteria, therefore, appears to be an important constraint on using them as
prescriptions of goals in political decision-making.
Several points should be taken into account when selecting criteria for ethical and/or political
decision-making. Firstly, the weighting factors presented here have considerable uncertainty
margins (as indicated, for example, by the limited consensus among the experts). Furthermore, it
must be noted that the criteria fall in different classes. Some most important criteria, such as AMI
and ‘tail and ear biting’, may be difficult to determine under practical conditions, as they would
either require direct (and long-term) observations of the animals or further modelling (which can
be done, for example, by drawing up lists of materials that have been shown to vary from low to
high AMI). For practical, policy-making purposes, therefore, it would be advised to select a
limited number of (the most important) feasible criteria where both the experts and the model
‘agree’. This study suggests that apart from the two main manipulation items (AMI and tail
biting) and the two ‘biologically functional objectives of AMI’ (namely explore/learn and
foraging), the following criteria may be considered as candidates for prescriptions of goals to
further implement EC directive 2001/93/EC in EC memberstates: novelty, rootability, bite/
chewability and changeabilty (i.e. destructibility). Keeping in mind the functional relationships
between these terms, e.g. destructible material is typically destructed because of biting/chewing
activity and requires renewal in order to maintain ‘destructibility’, two key criteria appear to be
rootability and destructibility. This is in accordance with the report of the SVC (1997), which
suggested that enrichment material for pigs should be either rootable (such as earth) or
destructible (such as straw).
5. Conclusions
We found a moderate correlation between the experts and the scores generated by the
preliminary model with respect to the importance of different criteria to assess environmental
enrichment materials. We also showed that general patterns might be identified explaining
M.B.M. Bracke et al. / Applied Animal Behaviour Science 104 (2007) 14–2322
differences between the model and the opinions of experts. It confirms that there is scope for
quantified assessment of animal welfare based on scientific information and suggests that there is
considerable scope for (further) modelling of environmental enrichment for pigs. The results
have been used for the further implementation of the EC directive in the Netherlands and may be
used in other EC memberstates, indicating that the results may be valuable for ethical and
political decision-making.
Acknowledgements
This work was funded by the Dutch Ministry of Agriculture, Nature and Food Quality
(Animal Welfare programme). We greatly acknowledge the contributions of the experts.
References
Anonymous, 1993. Genstat 5 Release 3 Reference Manual 1993. Oxford University Press, Oxford, 796 pp.
Anonymous, 2001. Scientists’ assessment of the impact of housing and management on animal welfare. J. Appl. Anim.
Welfare Sci. 4, 3–52.
Bracke, M.B.M., 2006. Semantic modelling: Evaluation of a formalised methodology for overall welfare assessment and
the assessment of enrichment materials for pigs, submitted for publication.
Bracke, M.B.M., Zonderland, J.J., Bleumer, E.J.B., 2007. Expert judgement on enrichment materials for pigs validates
preliminary RICHPIG Model. Appl. Anim. Behav. Sci. 104, 1–13.
Bracke, M.B.M., Zonderland, J.J., Lenskens, P., Schouten FW.G.P., Vermeer, H., Spoolder, H.A.M., Hendriks, H.J.M.,
Hopster, H, 2006. Formalised review of environmental enrichment for pigs in relation to political decision making.
Appl. Anim. Behav. Sci. 98, 165–182.
Bracke, M.B.M., Spruijt, B.M., Metz, J.H.M., Schouten, W.G.P., 2002a. Decision support system for overall welfare
assessment in pregnant sows A: model structure and weighting procedure. J. Anim. Sci. 8, 1819–1834.
Bracke, M.B.M., Metz, J.H.M., Spruijt, B.M., Schouten, W.G.P., 2002b. Decision support system for overall welfare
assessment in pregnant sows B: validation by expert opinion. J. Anim. Sci. 8, 1835–1845.
Bracke, M.B.M., Hulsegge, B., Keeling, L., Blokhuis, H.J., 2004. Decision support system with semantic model to assess
the risk of tail biting in pigs: 1. Modell. Appl. Anim. Behav. Sci. 87, 31–44.
Fraser, D., 2003. Assessing animal welfare at the farm and group level: the interplay of science and values. Anim. Welf.
12, 433–443.
Siegel, S., Castellan Jr., N.J., 1988. Nonparametric Statistics for the Behavioral Sciences, 2nd ed. McGraw-Hill, New
York.
SVC (Scientific Veterinary Committee), 1997. The welfare of intensively kept pigs. Report to the Directorate General
XXIV of the European Commission. Adopted September 30, 1997. Doc. XXIV/ScVc/0005/97, Scientific Veterinary
Committee, Animal Welfare Section, Brussels.
Van de Weerd, H.A., Docking, C.M., Day, J.E.L., Avery, P.J., Edwards, S.A., 2003. A systematic approach towards
developing environmental enrichment for pigs. Appl. Anim. Behav. Sci. 84, 101–118.
Young, R.J., 2003. Environmental Enrichment for Captive Animals. Blackwell, Oxford.
M.B.M. Bracke et al. / Applied Animal Behaviour Science 104 (2007) 14–23 23