Upload
andy-gray
View
90
Download
1
Embed Size (px)
Citation preview
Investigating the flexibility of selfish herd theory under
varying conditions in laying hens
A.D. GraySchool of Biological Sciences, Queen’s University Belfast, BT7 1NN, Northern Ireland
1
Table of contents
Abstract 3
Introduction 4
Method 10
Results 14
Discussion 19
Conclusion 21
Acknowledgements 22
References 23
2
AbstractThe selfish herd theory (Hamilton, 1971) is currently at the forefront of explaining the
psychology behind group herding behaviour during a time of perceived threat. In this
experiment - using selfish herd theory and subsequent literature as a basis - an artificial
predator (mini NSECT swarm: Tyco Mattel Inc.) was randomly presented to groups of
laying hens which were placed in either enriched (access to a foraging box) or unenriched
environments. The aim was to support the theory that predation would have an effect
upon clustering, and a secondary aim was to look at the effect of environmental condition
on clustering in four treatment conditions: ‘Predator’ and Foraging box present in pen
(PF); ‘Predator’ but no foraging box (P); Foraging box but no ‘predator’ (F); and no
‘Predator’ and no Foraging box (nPnF). Five groups of four hens were observed per
condition, giving a total of 20 groups across four conditions. Using video camera
observation, tracking software was used to plot the co-ordinates of the hens’ heads and
tails, in order to determine their level of clustering. The study found that the presence of
a predator had an effect upon clustering, with the birds clustering together during the
predation event, and returning to approximately the same distance apart after the event
that they were before. There was no significant effect of enrichment upon clustering. The
results are discussed with proposals for future research and the nature of the experiment
in an ethical context.
3
IntroductionSelfish herd theory is an area of science which is enjoying a new level of popularity. It is
considered one of the main contending theories on group herding, and much of the
pioneering work in this area was carried out in the 1970s, after which it was largely
ignored until the 1990s. Within the last 15 years, there has been a huge surge in the
amount of research that has been carried out, with grouping behaviour observed in a
range of animals including bees (Seeley and Buhrman, 1999), fish (Croft et al. 2003;
Stankowich and Blumstein, 2005), and birds (Hunter and Skinner, 1998; Couzin et al.
2002). It has also been observed in humans in a variety of different contexts, such as risk
from predators and environmental factors (Dyer et al. 2007; 2009; Brierley and Cox,
2010).
Selfish herd theory is based on a theoretical model which attempts to explain the
herding behaviour within groups in terms of benefits to the individual as opposed to the
group as a whole. The basis for the theory is a study by Hamilton (1971), who claimed
that the grouping together was actually caused by individual selfish motivations, as each
of the individuals tries to move into the geometrical centre of the group. By doing so,
individuals reduce the distance between themselves and others within the group, and thus
do not stand out as an individual and present a noticeable target to a predator. In
summary, ‘selfishness’ takes precedence over the welfare of others, and the combination
of these individual movements appears to be co-ordinated herding behaviour, but only
occurs due to a shared motivation to avoid predation.
Hamilton’s study itself was based on earlier work which had claimed that
companionship was the reason behind herd movements and their permutations (Galton,
1871). Hamilton attempted to examine all possible factors, such as the risk of predation
and self-preservation, and rejecting the common belief proposed by Galton of
companionship being the driving force behind herding behaviour.
There are other theories which attempt to explain herding behaviour. One is the
dilution effect, which states that individuals will cluster together to form a larger group,
and present an overwhelming number of potential targets to a predator, in order to
increase their own individual chance of survival (Foster and Treherne, 1981). While this
theory shares similarities to Hamilton’s theory, it is slightly flawed in that the defensive
4
strategy employed is dependent upon the attacking strategy of the predator, and thus is
susceptible to instability (such as another predator entering the fray, and causing the
dilution effect to fail). There is also an individual-based model, which assesses how
attraction and repulsion occurs between individuals with regard to herding (Gueron et al.
1996). If an individual moves into another individual’s comfort zone, the invaded
individual will move away to maintain an acceptable distance. Conversely, that
individual will also be attracted to another individual who is further away, and will move
closer to an acceptable distance. One of the explanations for this is the importance of
energetic costs; it appears to be an efficient strategy with regards to finding a balance
between individual vigilance and group vigilance. Yet the focus of the study is upon
individuals, rather than group herds. Due to the slight pitfalls in these theories, this
experiment will use Hamilton’s selfish herd theory as a basis for study and examination.
Several factors are integral to herding behaviour, one of the most important being
predation. Clearly it is an important factor; it is one of the main bases for selfish herd
behaviour, and its influence is fundamental, as has been found in the strategies of
predators in shaping fish schools during attacks (Parrish, 1992). Many past studies have
looked at the influence of predation; one particular study looked at the positioning of fish
in response to a chemical substance - known as ‘Schreckstoff’ - which is a chemical
substance located in the skin of minnows, and is released in response to a fearful
situation; in this case it is released in the presence of a predator. In this study, minnows
showed no particular structure with regards to spatial positioning, prior to the release of
Schreckstoff into the water. However, after its release - simultaneous with the predator
event - minnows tended to group together so that they were surrounded on all sides by
other fish (Krause, 1993). This follows the premise of the selfish herd theory - although
the cue to cluster is chemical, it demonstrates the effect of a predator upon clustering
behaviour by motivating individuals to make themselves less susceptible to predation by
clustering.
Prior to this study, the preference for forming as large a group as possible in the
presence of a predator had been observed (Hager and Helfman, 1991). In this study,
minnows were observed to prefer larger groups rather than smaller groups when a
largemouth bass was introduced, possibly indicating a show of strength in numbers.
5
However, this could simply be as a result of attempting to optimize group size and
dynamics, which has been observed in giant danio groups in which there are either too
few, or too many influential individuals (Viscido et al. 2005). In this situation, other fish
will ‘fan out’ to provide as much of a balance of influence within groups as possible - in
other words, to prevent a lopsidedness of influence and protect the whole of the group as
efficiently as possible. Again, the technique employed is based on selfish herd theory - to
present as many targets to a predator as possible, thus diminishing the threat to the
individual.
Grouping behaviour has been found in other species; such as fiddler crabs
(Viscido and Wethey, 2002). In this study, grouping behaviour is measured by a concept
known as the domain of danger. A domain of danger is the theoretical space around an
individual in a group in which that individual is susceptible to predation, and as a result
the individual will bunch together with other individuals (who have their own domains of
danger around them). This will therefore reduce the boundaries of the domain of danger -
due to proximity of other individuals - and reduce the overall threat towards the
individual.
Predation also influences selfish behaviour in birds (Quinn and Cresswell, 2007);
seals (de Vos and O’Riain, 2009); and wasps (Landi et al. 2002). The distance between
individuals and their grouping behaviour therefore appear to be important factors upon
defensive attempts to successfully avoiding predation. The impact of predation upon
clustering can be seen more clearly below (see Figure 1.):
6
Figure 1. Voronoi diagram of fiddler crab flock clustering before and after
predation threat (Viscido and Wethey, 2002).
In Figure 1, the effect of predation risk can be clearly seen upon clustering behaviour.
Before predation (Figure 1a), the crabs (represented by dots) are more spaced out, with a
number of small clusters. After predation (Figure 1b), the crabs form one large cluster,
with much less space around the individual. This figure sums up the net effect of the
selfish herd theory; selfish movements towards the centre of the group in order to remain
7
as safe as possible, avoid the periphery of the group, and thus present as small a target to
the predator.
With regards to the role played by predators during an attack, it has been found
that they are more likely to attack the periphery of a group, potentially indicating a
crucial choice factor for defensive individuals with regards to central group position
preference (Romey et al. 2008). This study was one of the first to study the effect of
predation in three dimensions, as opposed to previous aerial and 2-D dimensions in
previous studies, indicating its efficacy for real-life predation scenarios, and its reason for
inclusion in this study.
Computerised simulation models have found that an increase in predation risk
results in increased competition for central positions (which are deemed to be safer), and
during times of low predation, more foraging occurs and more peripheral positions are
preferred, as they are more useful for foraging due to the fact that their movements are
not restricted by being in the centre or surrounded by others (Morrell and Romey, 2008).
A similar study found that prey will actively choose between positions depending upon
the attacking/predation strategy of the predator, modifying their behaviour in response
(Morrell et al. 2010). With a predator being one our variables in this experiment,
predation is a factor of great importance. This study was followed up by another,
indicating that earlier models may not have the flexibility to incorporate the evolution of
predator strategy, and as such may become obsolete as an explanation of herding
behaviour (Morrell et al. 2011).
Another important factor with regards to which individuals achieve which
positions within the herd is dominance. Dominance is vital as it occurs within the group
at all times, including predation and non-predation times. During a predatory attack,
central positions within the herd are the most sought after and are generally won by more
dominant individuals due to their ‘safer’ nature (Hemelrijk, 2000; Morrell and Romey,
2008). The centre is not always safe however, as certain species such as black seabass
have been shown to attack the centre of a group more than the periphery (Parrish, 1989).
However, in another study individuals at the front of a moving group have generally been
found to be more susceptible to predation than those in central or rear positions (Bumann
et al. 1997). It seems therefore that the central position may be statistically the most
8
sought-after, but it is not proven to always be 100% safe in case of a predator attack. In
fact, it appears that a predation attack can be more random than prey can anticipate.
Predation and dominance are two of the main factors that can influence selfish
herd behaviour. However, they are not the only influences. The environment and
resources in which a group is habiting is crucial factor in herding behaviour, as has been
observed in laying hens. It has been found that solitary hens (hens that strayed from the
group) were much less rare at times of non-predation and clustering was found to occur
in response to the presence of environmental resources (Collins et al. 2011). This shows
evidence for collective herding behaviour and clustering outside times of predation, and
our study will attempt to discover if this system occurs during times of predation.
The need to forage, balanced against potential dangers in the environment, is also
an important factor. Hens have been found to prefer environments which induce lower
stress levels, namely those where they could forage (Nicol et al. 2009; 2011). The
presence of an area for foraging was another variable within our experiment.
The study described here used spatial co-ordinates of the hens recorded during
several conditions based on the above literature. Using computer-based programs, the
distances between each bird in a group were measured to calculate the level of clustering.
The aim of this experiment was to test the selfish herd theory (Hamilton, 1971) in a
domesticated species with no previous experience of predation, to determine whether an
individual’s position with regards to other birds in the group changes over time before,
during, and after a perceived predation threat, and whether environmental factors (such as
access to enrichment/foraging) have an effect upon clustering. With regards to the study
hypotheses, it was predicted that there would be an increase in clustering levels during
predation risk and a decrease before and after predation risk in line with past research
(Krause 1993; Viscido and Wethey, 2002). It was also predicted that there would be an
increase in clustering in enriched environments during times of predation and non-
predation (Nicol et al. 2009; 2011; Collins et al; 2011).
9
MethodSubjects and housingThe subjects consisted of 72 laying hens, which were of a commercially reared strain
(Hyline) and were approximately 18 weeks of age. Hens were housed in groups of four
within 18 wooden pens (96cm x 120cm in length and breadth, and 200cm high) at the
University of Bristol in the School of Veterinary Sciences between 2009-2010, and were
provided with feed from one large and one small feeding trough, which could be accessed
through openings at a height of 15 cm. A vertical water drinker was hung from the
ceiling at 20cm drinking height, and the floor was covered with wood shavings with two
perches placed across the width of the pen at 25cm in height and 0.35cm in diameter (see
Figure 2).
All birds were kept on a light schedule of 12 hours light and 12 hours darkness,
room temperature was kept between 18 and 22 degrees Celsius, and all birds were
sprayed with paint on their tails (Black, Red, White and Yellow) for easy identification
during the experiment, and which have been proven in past experiments to have no effect
on behaviour.
The hens were given two weeks to adjust to their new environment, and different
environments were then created to tie in with the four conditions, which were: ‘Predator’
and Foraging box present in the pen (PF); ‘Predator’ present, no Foraging box (P);
Foraging box present, no ‘Predator’ (F); and no ‘Predator’ and no Foraging box (nPnF).
For those conditions with a Foraging box present (PF and F), enrichment in the form of
an open-topped plastic box (35cm x 27cm in length and breadth, and 25cm high) for the
hens to forage in was placed in the middle of the pen to one side. For those conditions
without a foraging Box (P and nPnF), the box was closed and inaccessible to the birds,
though still present in the pen (see Figure 2).
For those conditions with a simulated ‘Predator’ (P and PF), a battery-powered
toy robot (mini NSECT swarm: Tyco Mattel Inc.) was used to simulate the predator. The
birds were exposed to this stimulus by opening the small feeder hatch, at which point the
robot moved its legs and flashed its eyes, thus representing a ‘Predator’, and causing an
avoidance response within the hens. The hatch remained open for a period of time
varying from 2 to 25 seconds before the hatch was replaced. For those conditions with
10
no predation risk (F and nPnF), the small hatch was never opened and the predator never
presented at any point.
Figure 2. Template of standard pen used in experiment. In half of the pens the
foraging box and drinker, and main feeder and small feeder, were on opposite sides
(Holt et al. submitted).
Data collectionVideo observation was used in this experiment, thus avoiding any potential effects of
audience bias. A video camera was attached to the ceiling of each pen, providing a birds-
eye view of the pen for observation. In total, 216 videos each of 8 hours in length were
collected by the researchers at the University of Bristol, which were then converted into
MPEG-4 format. For each of the four treatment conditions in this study, videos of five
unique groups of four hens were randomly selected, providing a total of 20 unique groups
across all treatments. Groups were chosen at random using the randomising
mathematical function in Microsoft Excel.
11
Selected videos were then imported into the Open Source Physics software Tracker
v3.16. This was used to record the head and tail co-ordinates (x and y) of each hen every
5 seconds, for 5 minutes before the predation risk event, during the predation risk event
itself (2-25 seconds), and for 5 minutes after the predation risk event. Therefore the
video for each group was just over 10 minutes in total observation time, though the exact
duration varied depending on the duration of the predation event. For those videos with
no predation, all videos were similar in length, with the start time chosen from the video
excerpts by once again using the randomising mathematical function in Microsoft Excel
(i.e. by randomly selecting a starting minute from each 8 hour video, excluding the final
10 minutes to avoid a premature cut-off in the final video excerpt). The time for each
‘during’ period (despite no ‘predator’ being present) was once again chosen randomly
using the method above, and was kept between 2-25 seconds.
After running the videos and recording the co-ordinates from Tracker, the x and
y coordinates for the head and tail of each bird in each group were imported into
Microsoft Excel. The inter-pair distances between all pairs within each group (Black-
Red, Black-Yellow, Black-White, Red-Yellow, Red-White, and Yellow-White) were
calculated for each recorded time-step (every 5 seconds during the record period). For
instance, calculating the distance between heads of the hens that were sprayed Black and
Red was done using the following geometrical formula:
√ (((Bhx – Rhx) x (Bhx – Rhx)) + (Bhy – Rhy) x (Bhy – Rhy)))
Bhx and Rhx represent the x coordinates for the head of the black-tailed and red-tailed
birds respectively, and Bhy and Rhy refer to the y coordinates. The same formula was
used when calculating the tail co-ordinates (with the x coordinates substituted for y
coordinates and subsequently written as Bty - Rty, and so on).
After calculating these distances for each pair of hens in the group, the average
distance between all pairs in the group was calculated for each 5-second time step as:
(( BR distance ) + ( BW distance ) + ( BY distance ) + ( RW distance ) + ( RY distance ) + ( WY distance )) 6
12
B, R, W, and Y are Black-, Red-, White- and Yellow-tailed birds, and the combination of
letters represents the pair of birds whose distance is being measured. Finally, an overall
average inter-pair distance for before, during, and after predation periods was calculated
by averaging across the average pair-distances per time-step for each of the three time
periods (averaging over 5 minutes pre-predator and post-predator, and over the duration
of the period that the predator was present).
The same technique was used in calculating average tail distances. Therefore, for
each of the 20 videos, there were six distance measures for statistical analysis; one
overall average group head distance and one overall average group tail distance for each
of the three time periods (before, during and after).
Statistical AnalysisIBM SPSS Statistics v.19 was used for statistical analysis. After following the step in the
previous section, there were six totals as mentioned above. Firstly, standardized residual
graphs of these data were created, in order to check for normality.
After finding that the data for clustering and predation did not fit the assumptions
for parametric testing, Wilcoxon signed rank tests were used to look at the within-group
differences in clustering between each of the three time periods (for both head and tail).
In order to look at the effect of predation risk on clustering in different groups in each
time period, an independent samples Mann-Whitney U test was used. This test was also
used for looking at the effect of predation on clustering between each time period for
head and tail. An independent samples Kruskal-Wallis test was used to look at the effect
of treatment condition upon both overall average head distance and overall average tail
distance.
All units of measurement in this study were in Tracker Units (TU). Parallax
transformations were not carried out because the camera looked straight down into the
pen, and as a result there were no differing angles between videos which could have
caused birds to appear different distances away. The results therefore are standardized as
Tracker Units were used as measurement in all videos.
13
ResultsAll conditionsThe difference between average head distance before and during (mean difference ± SE:
9.19TU ± 2.23TU) across all conditions was significant (Wilcoxon signed rank test,
p=0.04), as was the difference between average head distance during and after (mean
difference ± SE: 11.01TU ± 2.59TU) (Wilcoxon signed rank test, p=0.01). Before and
after (mean difference ± SE: 6.88TU ± 1.55TU) was not significantly different (Wilcoxon
signed rank test, p=0.601) (see Figure 3).
Figure 3. Absolute mean differences in overall group average head distance across
all time periods (see Table 1 for average head distances)
The differences between average tail distance before and during (mean difference ± SE:
12.64TU ± 2.55TU) (Wilcoxon signed rank test, p=0.079) and before and after (mean
difference ± SE: 9.11TU ± 1.63TU) (Wilcoxon signed rank test, p=0.526) were not
significant. However, the difference between during and after (mean difference ± SE:
11.85TU± 2.25TU) was significant (Wilcoxon signed rank test, p=0.03) (see Figure 4).
14
Figure 4. Absolute mean differences in overall group average tail distance across all
time periods (see Table 2 for average tail distances)
Predation conditionThis section covers those individuals in groups ‘P’ and ’PF’ only. The difference
between average head distance before (head mean before ± SE: 39.33TU ± 1.02TU) a
‘Predation’ event and during (head mean during ± SE: 32.36TU ± 0.67TU) was not
significant (Mann-Whitney U test, p=0.199). Similarly, the difference between average
head distance during and after (head mean after ± SE: 42.53TU ± 0.25TU) was not
significant (Mann-Whitney U test, p=0.821). However, the difference between average
head distance before and after was significant (Mann Whitney U test, p=0.041), with a
difference in average head distance of 3.16TU further after the event as compared to
before (see Figure 5).
15
Figure 5. Distance between heads in the predation condition across all time periods
The difference between average tail distance before (tail mean before ± SE: 41.14TU ±
1.93TU) and during (tail mean during ± SE: 33.41TU ± 0.58TU) was not significant
(Mann-Whitney U test, p=0.545). The difference between before and after (tail mean
after ± SE: 44.32TU ± 0.19TU) was not significant (Mann-Whitney U test, p=0.762), and
the difference between during and after was also not significant (Mann-Whitney U test,
p=0.496) (see Figure 6).
Figure 6. Distance between tails in the predation condition across all time period.
16
Foraging conditionThere was no difference in clustering between all four treatments (PF, P, F, nPnF). The
difference in average head distances before (mean ± SE: 43.65TU ± 0.34TU) and during
(mean ± SE: 39.70TU ± 1.65TU) the ‘predation’ event (in non-predatory treatments, this
was a randomly selected time period) was not statistically significant (Kruskal Wallis
test, p=0.361). Similarly the difference between before and after (mean ± SE: 43.25TU ±
0.47TU) across all treatments was not significant (Kruskal Wallis test, p=0.112), nor was
the difference between during and after (Kruskal Wallis test, p=0.687), showing no clear
influence of condition upon head distance (see Figure 7).
Figure 7. Distance between heads in the different foraging conditions (PF, P, F,
nPnF) across all time periods
Likewise, for the average tail distance, there was no difference across treatments in tail
distance differences before (mean ± SE: 48.16TU ± 0.28TU) and during (mean ± SE:
41.36TU ± 1.93TU) (Kruskal Wallis test, p=0.158), or between during and after (mean ±
SE: 46.60TU ± 0.53TU) (Kruskal Wallis test, p=0.217). However, there was a
statistically significant difference between before and after distances (Kruskal Wallis test,
p=0.044), with tail distance before 1.56TU further than after the predation event (see
Figure 8).
17
Figure 8. Distance between tails in the different foraging conditions (PF, P, F,
nPnF) across all time periods
Table 1. Average distance in overall head position across all time periods.
ConditionAverage distance
(Tracker Units)
Standard error
(Tracker Units)
PF (Head before) 42.56 0.96
P (Head before) 36.09 2.84
F (Head before) 44.74 0.95
nPnF (Head before) 50.13 3.24
PF (Head during) 34.50 1.91
P (Head during) 30.23 3.54
F (Head during) 44.90 1.70
nPnF (Head during) 38.53 2.20
PF (Head after) 41.75 1.23
P (Head after) 43.32 1.78
F (Head after) 44.74 1.10
nPnF (Head after) 52.30 2.34
18
Table 2. Average distance in overall tail position across all time periods
ConditionAverage distance
(Tracker Units)
Standard error
(Tracker Units)
PF (Tail before) 47.26 1.15
P (Tail before) 35.02 2.51
F (Tail before) 49.06 1.74
nPnF (Tail before) 48.68 2.71
PF (Tail during) 35.25 0.88
P (Tail during) 31.57 3.08
F (Tail during) 47.48 1.86
nPnF (Tail during) 45.58 5.15
PF (Tail after) 44.91 0.68
P (Tail after) 43.72 1.41
F (Tail after) 48.29 1.24
nPnF (Tail after) 49.01 1.95
DiscussionThe results indicate that overall, the hens were likely to stay a certain distance apart
before and after a perceived attack, while clustering more during a perceived attack. This
clearly supports the basis of the selfish herd theory (Hamilton, 1971). Notably the tails
were slightly further apart afterwards than before, whereas the heads were slightly close
together, perhaps emphasising the importance of keeping the head safe.
For predation, the same pattern emerged despite no statistical significance. Birds
still clustered together, but not to a significant degree. However they did return to a
statistically significant position, similar to that which they were in before. This indicates
that the birds roughly return to their previous domain keeping similar distances between
each other. This could indicate either a relatively quick recovery period or a slow
recovery period; a comparison with other species and their relative recovery times would
be an interesting area of research for the future. In addition, the fact that there was no
19
significance shown for tail distance indicates that birds appear to consider their head to be
more worthy of protection. This can also be supported by the larger distance between
bird tails before, during, and after the event, which roughly correlate with the differences
in head distances at the same times. Future studies could build on this idea, perhaps to
investigate if there is a preference or ‘formula’ for body position, or if the head is always
kept in the safest position.
The study also showed no effect of treatment condition overall, with no
significant difference in head distance throughout the time period across all conditions.
Therefore, it seems that hens appear to react the same to a predator attack when they have
access to enrichment/foraging box as when they do not.
The findings for predation therefore appear to support previous literature which
has discovered similar grouping behaviour by using the method of clustering by safety in
numbers (Hager and Helfman, 1991; Krause, 1993), and has also been found to support
other studies by reducing their domain of danger during times of predation (Viscido and
Wethey, 2002; Quinn and Cresswell, 2007; de Vos and O’Riain, 2009). The study also
supports the idea of clustering at times of foraging and non-foraging due to the relatively
small distances between the hens at times of non-predation – that is, before and after the
predation event (Collins et al, 2011). It is also of interest that the hens quickly return to
their previous positions, indicating a lack of post-event vigilance or trauma.
The study could also potentially support the idea for central group position
preference. Clustering occurs due to a movement towards the geometrical centre of the
group, and using a larger sample size of hens within a larger pen would be interesting in
order to look at the efficacy of the preference for central group position (Romey et al.
2008). A larger sample size may allow for support for the theory of more dominant birds
occupying central positions (Hemelrijk, 2000), and may also give support to the theory of
increased competition for central positions, combined with altering the periods and times
of predation (Morrell and Romey, 2008). In order to make these results more reliable, a
greater number of subjects within each condition could be beneficial (Nicol, 2007). The
fact that four birds were used in this experiment within each condition/pen limits the
scope of the conclusions that can be drawn.
20
It would also be interesting to build on this study by using a different predator other than
the mini NSECT swarm, or even an additional predator; this would allow for study upon
the birds changing of defensive positions with regards to the strategy of the attacker(s).
This could also be achieved by simply repeating the experiment, but altering the
predator’s position of exposure to the birds (i.e. making it appear from different points of
the pen, and not simply the small hatch each time). The predator could also be allowed
closer proximity to the birds, and it could thus be determined if the birds changed either
their position preference, or their clustering proximity (Morrell et al. 2010). The study
does show support for birds preferring less stressful environments; the fact that clustering
occurs at periods of heightened stress (during the attack) indicates that birds reside within
their own domains at times of relative peace and do not feel the need to gather in close
proximity (Nicol et al. 2009; 2011).
Our study showed some support for that of the individual-based approach
(Gueron et al. 1996). Birds did move away from each other after coming into close
proximity. Obviously they initially clustered due to their perceived need for safety, but
afterwards reduced their proximity, possibly due to achieving equilibrium with regards to
group vigilance or energetic costs. This could be an indicator of support for this
approach or for that of domains of danger (as mentioned earlier) or for both.
The study also has implications for animal welfare due to study of behavioural
responses to predator exposure. The fact that our hens showed clustering when visibly
exposed to a predator, could be used as a reason to implement new technology for
reducing visible exposure to external influences. As hens are not roaming animals, and
are happy to forage within their pens, normal hens and free range birds could be kept in
opaque housing, where potential predators such as foxes or larger birds are not visible to
the hens, and this in turn could prevent increased water loss due to heightened stress,
which has clear health implications (Puvadolpirod and Thaxton, 2000).
ConclusionWhat can be concluded from this paper therefore is that hens show implementation of
selfish herd actions when exposed to a predator, and that this occurs regardless of
environment. Therefore, the need to cluster in times of danger takes precedence over
21
other factors. However, with a limited predator attack such as is demonstrated in this
experiment, the extent to which we can place the level of importance upon this function is
also limited; many variables are missing that would be present in a real predator attack,
the main one being the lack of a real predator. As a real predator would more than likely
physically attack the birds and perhaps use a more cunning strategy, rather than merely
flash its eyes and move in a menacing way, the reaction of the birds and the interpretation
we can extrapolate from it is limited.
The ethical side of the experiment is paramount however. To implement such an
experiment which would produce truly reliable results would be unethical almost to the
point of being cruel, as only at a time of true predation would we be able to surmise
whether the selfish herd model is as rigid as it appears to be in this artificial setting.
However, we can conclude with certainty that clustering occurs at times of fright, and
that hens certainly feel safer in numbers than when on their own. We can also conclude
that peripheral positions are not preferred, as these are perceived even by the birds to be
unsafe. Therefore if this strategy is implemented at this relatively low level of
intelligence, we can surmise that its basic premises are utilised in more intelligent
species, such as cows, sheep etc. This study cannot prove selfish herd theory and
subsequent literature to be correct, but it also cannot disprove it, and in actual fact lends
support to the theory, thus maintaining its position as the foremost explanation behind
herding behaviour in times of danger.
AcknowledgementsI would like to acknowledge the work carried out by Christine Nicol and her team of
researchers at the University of Bristol, who collected all video footage used in this
experiment and without which this study would not have been possible. I would also like
to thank Hannah Holt of the Royal Veterinary College, London, for her invaluable help
and technical advice.
22
ReferencesBrierley, A.S., and Cox, M.J. 2010. “Shapes of Krill Swarms and Fish Schools Emerge
as Aggregation Members Avoid Predators and Access Oxygen.” Current
Biology, 20 (19), 1758-1762.
Bumann, D., Krause, J., and Rubenstein, D. 1997. “Mortality risk of spatial positions
in animal groups: the danger of being in the front.” Behaviour, 134 (13/14),
1063-1076.
Collins, L.M., Asher, L., Pfeiffer, D.U., Browne, W.J., and Nicol, C.J. 2011.
“Clustering and synchrony in laying hens.” Applied Animal Behaviour Science,
129, 43-53.
De Vos, A., and O’Riain, M.J. 2009. “Sharks shape the geometry of a selfish seal herd:
experimental evidence from seal decoys.” Biology Letters, 6, 48-50.
Foster, W.A., and Treherne, J.E. 1981. “Evidence for the dilution effect in the selfish
herd from fish predation on a marine insect.” Nature, 293, 466-467.
Galton, F. 1871. “Gregariousness in cattle and men.” Macmillan’s Magazine, London,
23, 353.
Gueron, S., Levin, S.A., and Rubenstein, D.I. 1996. “The Dynamics of Herds: From
Individuals to Aggregations.” Journal of Theoretical Biology, 182, 85-98.
Hager, M.C., and Helfman, G.S. 1991. “Safety in numbers: shoal size choice by
minnows under predatory threat.” Behavioural Ecology and Sociobiology, 29,
271-276.
Hamilton, W.D. 1971. “Geometry for the Selfish Herd.” Journal of Theoretical
Biology, 31, 295-311.
Hemelrijk, C. 2000. “Towards the integration of social dominance and spatial
structure.” Animal Behaviour, 59, 1035-1048.
Holt, H., Collins, L.M., Asher, L., Browne, W., Pfeiffer, D.U., Caplan, G., Statham,
P., and Nicol, C. 2011. “Fractal analysis of laying hens’ movements:
associations with environment, behaviour and physical condition.” (submitted)
Krause, J. 1993. “The effect of ‘Schreckstoff’ on the shoaling behaviour of the
minnow: a test of Hamilton’s selfish herd theory.” Animal Behaviour, 45, 1019-
1024.
23
Landi, M., Coster-Longman, C., and Turilazzi, C. 2002. “Are the selfish herd and the
dilution effects important in promoting nest clustering in the hover wasp
Parischnogaster alternata (Stenogastrinae Vespidae Hymenoptera)?” Ethology,
Ecology and Evolution, 14, 297-305.
Morrell, J.L., and Romey, W.L. 2008. “Optimal individual positions within animal
groups.” Behavioural Ecology, 19, 909-919.
Morrell, J.L., Ruxton, G.D., and James, R. 2010. “Spatial positioning in the selfish
herd.” Behavioural Ecology, 21, 1367-1373.
Morrell, J.L., Ruxton, G.D., and James, R. 2011. “The temporal selfish herd:
predation risks while aggregations form.” Proceedings of the Royal Society of
Biological Sciences, 278, 605-612.
Nicol, C.J. 2007. “Space, time, and unassuming animals.” Journal of Veterinary
Behaviour, Clinical Applications and Research, 2 (6), 188-192.
Nicol, C.J., Caplen, G., Edgar, J. and Browne, W.J. 2009. “Associations between
welfare indicators and environmental choice in laying hens.” Animal Behaviour,
78, 413-424.
Nicol, C.J., Caplen, G., Statham, P., and Browne, W.J. 2011. “Decisions about
foraging and risk trade-offs in chickens are associated with somatic response
profiles.” Animal Behaviour, 82 (2), 255-262.
Parrish, J.K. 1989. “Re-examining the selfish herd: are central fish safer?” Animal
Behaviour, 38 (6), 1048-1053.
Parrish, J.K. 1992. “Do predators ‘shape’ fish schools: interactions between predators
and their schooling prey.” Netherlands Journal of Zoology, 42 (2-3), 358-370.
Puvadolpirod, S., and Thaxton, J. 2000. “Model of physiological stress in chickens 4.
Digestion and metabolism.” Poultry Science, 79, 383-390.
Quinn, J.L., and Cresswell, W. 2007. “Testing domains of danger in the selfish herd:
sparrowhawks target widely spaced redshanks in flocks.” Proceedings of the
Royal Society of Biological Sciences, 273, 2521-2526.
Romey, W.L., Walston, A.R., and Watt, P.J. 2008. “Do 3-D predators attack the
margins of 2-D selfish herds?” Behavioural Ecology, 19, 74-78.
24
Viscido, S.V., and Wethey, D.S. 2002. “Quantitative analysis of fiddler crab flock
movement: evidence for ‘selfish herd’ behaviour.” Animal Behaviour, 63, 735-
741.
Viscido, S.V., Parrish, J.K., and Grunbaum, D. 2005. “The effect of population size
and number of influential neighbours on the emergent properties of fish schools.”
Ecological Modelling, 183 (2-3), 347-363.
25