See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/250719206

Application of a fully automatic analysis tool to assess the activity of broiler

chickens with different gait scores

Article in Computers and Electronics in Agriculture · August 2010

DOI: 10.1016/j.compag.2010.05.004

CITATIONS READS

45 276

5 authors, including:

Arda Aydin C. Bahr

Çanakkale Onsekiz Mart Üniversitesi KU Leuven
20 PUBLICATIONS 82 CITATIONS 97 PUBLICATIONS 944 CITATIONS

SEE PROFILE SEE PROFILE

Daniel Berckmans
KU Leuven
514 PUBLICATIONS 4,576 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

BioBusiness View project

ammonia measurement View project

All content following this page was uploaded by Arda Aydin on 06 July 2018.

The user has requested enhancement of the downloaded file.

 Computers and Electronics in Agriculture 73 (2010) 194–199

Contents lists available at ScienceDirect

Computers and Electronics in Agriculture

journal homepage: www.elsevier.com/locate/compag

Application of a fully automatic analysis tool to assess the activity of broiler

chickens with different gait scores
A. Aydin, O. Cangar, S. Eren Ozcan, C. Bahr, D. Berckmans ∗
Division Measure, Model & Manage Bioresponses, Katholieke Universiteit Leuven, Kasteelpark Arenberg 30, B-3001 Heverlee, Belgium

a r t i c l e i n f o a b s t r a c t

Article history: Monitoring the activity of broilers is a potential way for determining gait score level at commercial
Received 4 January 2010 farms. In this study, a fully automatic monitoring technique was developed to measure the activity of
Received in revised form 31 March 2010 broiler chickens with different gait score levels. We carried out two experiments in order to assess the
Accepted 9 May 2010
relationship between gait scores obtained by human experts and activity levels quantified by an automatic
image monitoring system. The chickens were scored for their degree of lameness by experts according to
Keywords:
the method of Kestin et al. (1992). For each experiment, a total of 30 birds, on day 32 were selected from
Activity
a local commercial farm. Five birds were selected in each of six gait score groups (GS0–GS5). The activity
Broiler chickens
Gait score
levels were obtained for all gait scores by using an automatic image monitoring system. For this purpose,
Automatic monitoring video surveillance images of broilers with six different predefined gait scores were analysed. In both
experiments, there was a significant relation between gait score by experts and activity monitored by
image analysis. The broilers with gait score 3 (GS3) showed significantly higher activities than the other
gait scores in both experiments (P < 0.05), possibly due to their need for more feed. In both experiment,
GS4 and GS5 showed significantly lower activities (P < 0.05). Overall, the results show that automatic
camera monitoring system can provide an automatic tool in determination the activity in relation to
gait score. This activity information can be used further to identify the effects of gait score on broiler
behaviour.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction al., 1997). High light intensity increased activity of broiler chick-
ens and decreased leg problems and mortality without affecting
The most important questions related with broilers welfare production (Cherry and Barwick, 1962). Rearing broilers in bright
quoted in the last two decades are the growing sensitivity for red light early in life increased activity and decreased leg disor-
metabolic and locomotion problems due to the fast growth rates ders compared with broilers reared in dim blue light (Prayitno et
and inactivity of the chickens. In commercial farm conditions, al., 1997), whilst environmental complexity has been reported to
broiler chickens show low levels of activity, particularly the ones either increase activity (Bokkers and Koene, 2003) or have no effect
kept in high stocking densities and during the last weeks of the upon activity or gait score (Bizeray et al., 2002).
growing period (Blokhuis and Van Der Haar, 1990; Bizeray et al., Traditional methods for determination of gait score include the
2002). There are many potential welfare benefits of monitoring the manual scoring of animal behaviour in the broiler house. Recorded
activity of broiler chickens in the farm. Increased or decreased over- images can also be used for manual scoring of chicken gait score. But
all movement within the flock may imply an emerging disease or scoring of some animal-based information by human experts and
change in behaviour towards aggression due to lack of feed or water manual methods remain difficult, time consuming and expensive
(Swayne and Halvorson, 2003). when implemented at farm level.
Several studies have focused upon ways of increasing activ- Image analysis technologies have been widely used in behaviour
ity early in the growing period, although reducing activity is analysis of different animals. Thermal comfort behaviour of swine
sometimes desirable, for example during decreasing of popula- was analyzed by Shao et al. (1998) using programmable cameras.
tion (Bizeray et al., 2002; Bokkers and Koene, 2003; Prayitno et The area and the perimeter of the top view of the pigs could be
extracted from the images. Individual behaviour of pigs in a pen
was studied by Tillett et al. (1997). In their work, an image pro-
cessing technique was used to track the animal movements. The
∗ Corresponding author at: Division Measure, Model & Manage Bioresponses,
fitting of a model to the top view image sequence provided data on
Katholieke Universiteit Leuven, Kasteelpark Arenberg 30, B-3001 Leuven, Belgium.
Tel.: +32 0 16321726; fax: +32 0 16321480.
position, rotation, bending and head nodding. The locomotion and
E-mail address: Daniel.berckmans@biw.kuleuven.be (D. Berckmans). posture behaviour of pregnant cows prior to calving was studied by

0168-1699/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.compag.2010.05.004
 A. Aydin et al. / Computers and Electronics in Agriculture 73 (2010) 194–199 195

Cangar et al. (2008). In their study, automatic real-time monitoring recover from the stress of transport and acclimatize to their new
system was used to classify specific behaviours such as standing environment. Lights were kept on during the video recordings as
or lying (including incidences of motion during lying), and eating shown in Table 1.
or drinking. Similar studies have also been performed to detect of A digital video camera, Guppy F036C equipped with a C30811KP
lameness in dairy cattle (Song et al., 2008; Tasch and Rajkondawar, 8.5 mm lens (Pentax) was mounted 4.1 m above the floor with its
2004). Leroy et al. (2005) established a model-based computer lens pointing downwards and directly above the center of the six
vision system to study the behaviour of hens in furnished cages. pens to get a top view of all pens in the camera image (see Fig. 1b).
Individual behaviours such as standing, walking and scratching The camera was connected to a PC with a built-in frame grabber
could be recognised automatically and in real time. (E119932-U, AWM 20276, VW-1) using an IEEE 1394 fire wire cable.
Using video camera images for analysing activity is becoming Images were captured with a resolution of 1024 × 768 pixels at a
an emerging technology. It is relatively cheap and non-invasive sample rate of 3.5 frames per second. Video recordings were made
technique that facilitates the collection of more frequent data over during 5 days.
longer time periods. In case of processing, an analysing algorithm
in real time is used. Therefore, no huge data storage is required.
The existing image analysis tools were developed in pig and 2.2. Image calibration
cow chambers in laboratory conditions. In commercial livestock
houses, image analysis for behaviour classification becomes more Prior to the experiments, the image was calibrated so that the
complicated. Lighting, camera characteristics, background and test areas of pixels in the image could be converted to units of cm2 on the
subject’s traits all influence the ability of the system to recognise pen floor. With the known dimensions of the pen (1 m × 1 m) and by
the subject and record its movement accurately (Hoy et al., 1996). measuring these distances in the camera pixels in units of pixels,
A new technique was developed by Sergeant et al. (1998) to derive a linear factor relating image coordinates to positions within the
a background image representing the scene without the objects of broiler pen could be estimated as f = 0.33 cm per pixel. Therefore,
interest and perform frame by frame image subtraction for com- the distance between two points, one pixel apart is 0.33 cm on the
puter visual tracking of poultry. A new hypothesis that measures pen floor and the area of a region the size of one pixel is f2 = 0.11 cm2
valuable on-farm outcome of broiler (meat) chicken welfare can be on the pen floor.
derived from optical flow statistics of flock movements recorded
on video or CCTV inside commercial broiler houses was tested by
Dawkins et al. (2009). 2.3. Activity measurements
The objective of this study is to investigate the activity levels of
broiler chickens in relation to their gait scores using an automatic Activity of different gait score chickens was measured using the
image monitoring system under laboratory conditions. Further- “Eyenamic software” (Leroy et al., 2006). The software automati-
more, a fully automatic monitoring and image analysis system was cally grabbed 3.5 frame per second monochrome images I(x, y, t)
applied to determine the average activity levels of chickens with from the camera, with I being the intensity of the pixel at coordi-
different gait scores. The outcome of this study serves as a prelim- nates (x, y) in that image. The difference of the intensity values with
inary step for developing an automatic behaviour analysis tool for the previous image I(x, y, t − 1), was calculated. From this difference
chickens with different gait scores in commercial farms. image, the binary ‘activity image’ Ia (x, y, and t) was calculated,
containing the pixels for which the intensity change exceeded a
2. Materials and methods threshold:

2.1. Birds, experimental design and video recordings 1 if I(x, y, t) − I(x, y, t − 1) > 1
I a (x, y, t) = (a)
0 otherwise
Two experiments were conducted during 5 days in two differ-
ent years. In experiment 1 which was conducted in 2008, 14.68 h From the activity image Ia (x, y, t) the activity index ai (t) for
of data were recorded. In experiment 2 which was conducted in zone Zi was calculated as the fraction of moving pixels with respect
2009, 51.58 h of data were recorded. The second experiment is a to the total number of chicken area pixels (1). Activity image (Ia )
repetition of the first experiment except of the recording times. area was normalised by the total area of the chickens (1) in each
The experiments were conducted with Ross 308 broilers which compartment to compare the results independent of the size of the
were vaccinated at the hatchery and on day 23 in the stables, fol- chickens.
lowing standard procedure. For the first 9 days, a prestarter diet

with 23% protein and 2890 kcal AMEn/kg (apparent metabolisable I (x, y, t)
(x,y) ∈ Zi a
energy) was given. From day 10 until day 13, starter diet with 22% ai (t) =  (b)
(x,y) ∈ Zi
1
protein and 2794 kcal AMEn/kg, and from day 14 to day 32, grower
diet with 20% protein and 2899 kcal AMEn/kg, were provided.
A total of 60 birds, mixed sex, 32 days old were selected from To eliminate errors occurring due to different size of birds, the
a local commercial farm (Provincial Center for Applied Poultry total amount of measured movement was normalised by divid-
Research, Geel, Belgium) using the gait scoring method of Kestin et ing the average size of the birds in each pen. The threshold  1
al. (1992). According to Kestin et al. (1992), lameness of the chick- accounted for small intensity changes due to noise, e.g. electrical
ens were ranked in the increasing order from gait score zero (GS0) noise in the coax cabling and image acquisition circuits, small light-
to gait score five (GS5) where GS0 is being the healthiest. The birds ing variations, etc. Recordings in each experiment had a different
then were transported to the laboratory. background lighting condition. Therefore the value of the threshold
The laboratory test installation had six stainless steel compart- was set to 10% of the maximal intensity in each specific recording
ments (100 cm × 100 cm, width × length). In each compartment the separately. It was estimated manually by looking at the image data
same number of mixed sex birds having the same gait score level since no reference values from previous studies have been avail-
was housed with a stocking density of 5 birds/m2 . Birds were kept able. The pixel area summed in the nominator and denominator of
on floor pens on wood shavings. Feed and water were freely avail- Eq. (b) has an accuracy of one pixel which was equivalent to an area
able to all birds. Birds were allowed a minimum of two days to of 0.11 cm2 using the camera calibration factor was.
196 A. Aydin et al. / Computers and Electronics in Agriculture 73 (2010) 194–199

Table 1
Daily numbers of frames and recording times of the experiments.

Year Experimental day Recording times Number of frames Year Experimental day Recording times Number of frames

1 00:35:01 113523 1 21:33:10 289796

2 00:02:57 961 2 04:45:08 62998
2008
3 02:19:09 83496 2009 3 00:56:52 14400
4 02:46:40 961 4 20:29:54 276298
Total 05:43:47 198941 5 02:25:37 32400
Total 50:10:41 675892

2.4. Statistical analysis dencies do not affect the test results. Sample size was decreased
by cumulating 1050 measurements to one activity value per every
Friedman’s Test was used to analyze the effects of gait score on five minutes of recordings. This was done because otherwise there
birds’ activity. Friedman’s Test is a non-parametric test that com- were more than 180,000 samples for each experiment which was
pares the columns without the row effects. Therefore, it does not impossible to analyze with the mentioned statistical tests. After
test for row or interaction effects. In the test sample size and depen- the Friedman’s Test, Dunn’s Test was used to define the statistical

Fig. 1. Laboratory set up with a computer, camera, cages with five chickens in each compartment (a), and one example from recorded top view images (b).
 A. Aydin et al. / Computers and Electronics in Agriculture 73 (2010) 194–199 197

differences between the gait scores. The calculations were per- As can be seen in Table 2, the results of the Friedman’s Test
formed using the Statistics Toolbox of Matlab (The Math Works, revealed that there is a significant relation between the gait score
Massachusetts, USA). and activity in both experiments (see Table 2, P < 0.05).
As shown in Table 3, activity of GS3 (mean ± standard devia-
tion) was found significantly higher than the other gait scores in
the first experiment (4.82 ± 3.40b), as opposed to our expectations.
3. Results
Although there is an overall correlation between activity score and
gait score, this was not linear and GS3 has the highest activity.
3.1. Activity and gait score
Table 3 shows that there is no significant difference between GS0,
GS1 and GS2 (see Fig. 3a, b). The authors claim that impairment
A fully automatic, image monitoring tool was applied to calcu-
of movement of GS0 to GS3 chickens depends on their increasing
late the activity index of in total 30 chickens divided into six groups
weight while GS4 and GS5 are rather clinically sick (lame) chickens.
with a different gait score.
One of the possible reasons would be that GS3 chickens are big-
As it can be seen in Fig. 2, there is a permanent change in activ-
ger and heavier than the other groups (see Table 3); therefore they
ity over time. No specific pattern has been observed. Therefore,
peck and fight for feed more than the other groups. In this group
cumulative activity values over time have been taken to show the
broiler’s need to eat was higher than the probable discomfort they
differences between the gait scores clearly, since the time series
might have had by presenting an abnormal gait score. As it can
data was very noisy to interpret efficiently (see Fig. 2a).
be seen in Fig. 4 and Table 3, there was a significant correlation
Fig. 3 shows daily cumulative activity levels at the end of each
between weight of the birds and their gait score especially for lamer
experiment. Since the recording times were not equal between the
birds (P < 0.05). GS3 chickens were heavier and showed the highest
days, cumulative activity of different gait score chickens reached
activity in both trials. Besides GS4&GS5 were significantly lighter
different values at the end of the experiments. By dividing the
and less active compared to other chickens.
frames into five minutes intervals, it was possible to compare sub-
In the second experiment, the highest activity was again found
cumulative activities in both experiments with different recording
in GS3 (6.08 ± 4.66b) as in the first experiment (P < 0.05). Similar to
times (Table 1).

Fig. 2. One example of continuous activity index measurement of GS3 group chickens during 3-h experiment 1 (a) and experiment 2 (b).

Fig. 3. Activity index of the different gait score chickens during experiment 1 (a) and experiment 2 (b).
198 A. Aydin et al. / Computers and Electronics in Agriculture 73 (2010) 194–199

Table 2
Results of Analysis of Variance for the effects of gait score on activity.

Source (a)Experiment 1

Sum of squares Degrees of freedom Mean squared Chi-sq P(>Chi-sq)

Gait score 1148.90 5 229.78 328.26 0

Error 15336.10 4705 3.26
Total 16485 5651

Source (b)Experiment 2

Sum of squares Degrees of freedom Mean squared Chi-sq P(>Chi-sq)

Gait score 2324.53 5 464.91 664.15 0

Error 8700.47 3145 2.77
Total 11025 3779

Table 3
The estimated mean activity index for broiler chickens at different gait scores with standard deviation, weight and body area of chickens.

Activity index Weight (kg) Body area (cm2 )

4 5
Exp. 1 (×10 ) Exp. 2 (×10 ) Exp. 1 Exp. 2 Exp. 1 Exp. 2

GS0 1.94 ± 1.451a 4.51 ± 3.831a 1.16 ± 0.281ab 1.32 ± 0.32a 208.06 ± 42.291ab 235.66 ± 47.881ac
GS1 2.28 ± 2.39a 4.27 ± 4.07a 1.35 ± 0.15a 1.49 ± 0.17ab 231.79 ± 23.37ab 209.81 ± 21.18ac
GS2 2.62 ± 2.17a 5.02 ± 4.57b 1.31 ± 0.09a 1.67 ± 0.12ab 280.63 ± 16.24b 220.43 ± 12.78ab
GS3 4.82 ± 3.40b 6.08 ± 4.66b 1.45 ± 0.05a 1.76 ± 0.13b 307.42 ± 19.19b 227.42 ± 6.81b
GS4 2.19 ± 2.07c 1.58 ± 1.37c 1.28 ± 0.30a 1.27 ± 0.30a 185.24 ± 40.21ab 187.16 ± 40.62c
GS5 0.98 ± 0.90d 1.89 ± 1.68d 0.90 ± 0.10b 1.37 ± 0.16ab 245.38 ± 23.20a 160.70 ± 15.20a
1
Mean ± Standard Deviation.
a-b-c
means, within a column, with no superscript in common differ significantly (P < 0.05).

the first experiment, GS3 chickens were again bigger and heavier of different gait score chickens may indeed be different in other
(see Table 2). Therefore, it is assumed that they frequently visited breeds.
the feeder. However, GS4 (1.58 ± 1.37c) and GS5 (1.89 ± 1.68c) were
found to have the lowest activities in the second experiment like
as the first experiment. As it can be seen in Table 3, there was no 4. Discussions and conclusions
significant difference in the level of activity between the GS0 and
GS1 in both experiments. Studies suggest that broilers are motivated to walk long dis-
Cumulative activity differences between the gait scores over tances for feed and that their motivation can be manipulated, also
three hour measurement can be seen in Fig. 5. Especially, signifi- within body weight groups. Accelerated growth rates and heavier
cant activity difference between the GS4–GS5 and other gait scores body weights were stated to have an influence on the locomo-
could be easily seen in the second experiment in Fig. 5b. The value of tion (Kestin et al., 2001). A heavy body weight requires more
standard deviation in all groups was found very high in both exper- from the yet not fully grown skeleton system and that leads to
iments. This is due to the behavioural differences of each individual abnormal ‘gait scores’ (Corr et al., 2003). Lameness significantly
bird. changes the time budgets of much behaviour and dramatically
Since in both experiment, the same broiler breed Ross 308 was alters feeding strategy (Weeks et al., 2000). The nature of the
used. These results above inform only the behaviour of Ross 308 apparent relationship between lameness and reduced activity lev-
as the most common breed in Europe. The activity index results els remains unclear in literature (Hester, 1994). In this study, the

Fig. 4. Correlation between the gait score and weight of the chickens.
 A. Aydin et al. / Computers and Electronics in Agriculture 73 (2010) 194–199 199

Fig. 5. One example of the cumulative activity index of different gait score chickens during experiment 1 (a) and experiment 2 (b), over three hours.

relation between the gait score and activity was investigated in ing durations of feed deprivation and feed access. Appl. Anim. Behav. Sci. 108,
60 commercial broiler chickens reared in laboratory conditions for 129–142.
Cangar, O., Leroy, T., Guarino, M., Vranken, E., Fallon, R., Lenehan, J., Meed, J., Berck-
two experiments. It was found that a significant relation between mans, D., 2008. Automatic real-time monitoring of locomotion and posture
the gait scores and activity exists (p < 0.05). In contrast to our behaviour of pregnant cows prior to calving using online image analysis. Com-
expectations, the relation between activity and gait score was not put. Electron. Agric. 64, 53–60.
Cherry, P., Barwick, M.W., 1962. The effect of light on broiler growth. I. Light intensity
linear. The GS3 birds reached the highest activity level instead and colour. Br. Poult. Sci. 3, 31–39.
of GS0 (Fig. 1). Bokkers et al. (2007) showed that the high body Corr, S.A., Gentle, M.J., Mccorquodale, C.C., Bennett, D., 2003. The effect of morphol-
weight of broilers can be considered as a physical constraint to ogy on walking ability in the modern broiler: a gait analysis study. Anim. Welf.
12 (2), 159–171.
be active and probably to behave normally. However, as shown Dawkins, M.S., Lee, H.J., Waitt, C.D., Roberts, S.J., 2009. Optical flow patterns in broiler
in Table 3, we found that the most active GS3 chickens had the chicken flocks as automated measures of behaviour and gait. Appl. Anim. Behav.
highest body area (307.42 ± 19.19b cm2 ) during the experiments. Sci. 119 (3–4), 203–209.
Hester, P.Y., 1994. The role of environment and management on leg abnormalities
This could be explained by the weight of the birds. As can be seen
in meat-type fowl. Poult. Sci. 73, 904–915.
in Table 3, the body weights of chickens are significantly differ- Hoy, J.B., Koehler, P.G., Patterson, R.S., 1996. A microcomputer-based system for
ent (P < 0.05) and the GS3 chickens had the highest body weight real-time analysis of animal movement. J. Neurosci. Methods 64, 157–161.
(1.76 ± 0.13b kg). The higher activity might be explained by a higher Kestin, S.C., Knowles, T.G., Tinch, A.E., Gregory, N.G., 1992. Prevalence of leg weakness
in broiler chickens and its relationship with genotype. Vet. Rec. 131, 190–194.
need of feed. They need more feeding than the other birds, and Kestin, S.C., Gordon, S., Su, G., Sorensen, P., 2001. Relationships in broiler chickens
they could be more active than the others because they eat more between lameness, live weight. Growth Rate Age Vet. Rec. 148 (7), 195–197.
although in this study, the feed intake of chickens has not been Leroy, T., Vranken, E., Van Brecht, A., Struelens, E., Janssen, A., Tuyttens, F., De Baere,
K., Zoons, J., Sonck, B., Berckmans, D., 2005. A quantitative computer vision
quantified. method for on-line classification of poultry behavior in furnished cages. Trans.
In this research it is stated that there is a significant relation ASAE 49 (3), 795–802.
between the gait scores and activity (p < 0.05), and more experi- Leroy, T., Vranken, E., Van Brecht, A., Struelens, E., Sonck, B., Berckmans, D., 2006.
A computer vision method for on-line behavioral quantification of individually
ments are needed to see the repeatability of the results. caged poultry. Trans. ASABE 49 (3), 795–802.
This automatic monitoring of activity will allow researcher to Prayitno, D.S., Phillips, C.J.C., Stokes, D.K., 1997. The effects of color and intensity of
study behaviour analysis of different gait score groups. Therefore, light on behavior and leg disorders in broiler chickens. Poult. Sci. 76, 1674–1681.
Sergeant, D., Boyle, R., Forbes, M., 1998. Computer visual tracking of poultry. Comput.
in future, implementation of suggested system in field conditions Electron. Agric. 21, 1–18.
should be tested. These experiments should be repeated to observe Shao, J., Xin, H., Harmon, J.D., 1998. Comparison of image feature extraction for
whether the findings of experiment 2 is consistent, that is, there is classification of swine thermal comfort behavior. Comput. Electron. Agric. 19
(3), 223–232.
a significant low activity of high gait score chickens (GS4&GS5). In
Song, X., Leroy, T., Vranken, E., Maertens, W., Sonck, B., Berckmans, D., 2008. Auto-
that case, this automatic activity monitoring tool can be used on matic detection of lameness in dairy cattle—vision-based trackway analysis in
indicator of high gait scores (GS4&GS5) in the field conditions. cow’s locomotion. Comput. Electron. Agric. 64 (1), 39–44.
Swayne, D.E., Halvorson, D.A., 2003. Influenza. In: Diseases of Poultry, vol. 147., 11th
ed. Iowa State University Press, Ames, IA, 1997.
References Tasch, U., Rajkondawar, P.G., 2004. The development of a SoftSeparator (TM) for a
lameness diagnostic system. Comput. Electron. Agric. 44 (3), 239–245.
Bizeray, D., Estevez, I., Leterrier, C., Faure, J.M., 2002. Effects of increasing envi- Tillett, R.D., Onyango, C.M., Marchant, J.A., 1997. Using model-based image process-
ronmental complexity on the physical activity of broiler chickens. Appl. Anim. ing to track animal movements: livestock monitoring. Comput. Electron. Agric.
Behav. Sci. 79, 27–41. 17 (2), 249–261.
Blokhuis, H.J., Van Der Haar, J.W., 1990. The effect of the stocking density on the Weeks, C.A., Danbury, T.D., Davies, H.C., Hunt, P., Kestin, S.C., 2000. The behaviour
behaviour of broilers. ZADI (Germany, Federal Republic of) 54 (2), 74–77. of broiler chickens and its modification by lameness. Applied Animal Behaviour
Bokkers, E.A.M., Koene, P., 2003. Behaviour of fast- and slow-growing broilers to 12 Science, vol. 67. Department of Clinical Veterinary Science, University of Bristol,
weeks of age and the physical consequences. Appl. Anim. Behav. Sci. 81, 59–72. Langford, Bristol BS40 5DU, UK, pp. 111–125.
Bokkers, E.A.M., Zimmerman, P.H., Bas Rodenburg, T., Koene, P., 2007. Walking
behaviour of heavy and light broilers in an operant runway test with vary-

View publication stats