Behavior of Broiler Chickens in Four Different Substrates : a Choice Test

I Centro de Tecnología Animal IVIA. Polígono La Esperanza no100, 12400, Segorbe, Castellón, Spain II Cátedra de Biofísica, Departamento de Ciencias Básicas, Facultad de Ciencias Veterinarias, Universidad Nacional del Litoral, R. P. L. Kreder 2805 (3080) Esperanza, Santa Fe, Argentina III Instituto de Ciencia y Tecnología Animal, Universidad Politécnica de Valencia, Camino de Vera s/n, 46022, Valencia, Spain

In recent years, materials like sawdust or wood shavings have become scarce, due to the fact that they are used as fuel, so the broiler industry has had to look for new and readily available materials.In the search for new materials for broiler litter, researchers must seek materials that meet hygienic requirements and guarantee that ammonia levels do not exceed certain levels throughout the productive cycle (Worley et al., 1999).In this sense, distinct mixtures of different materials have been proposed as bedding for poultry (e.g.Sanotra et al., 1995;Willis et al., 1997;Al Homidan & Robertson, 2003).The addition of other products to improve bird performance and litter characteristics is also being investigated (e.g.natural zeolite in Eleroglu & Yalcin, 2005; alum in Worley et al., 1999).
Moreover, the characteristics of the materials used as broiler substrate must be taken into account, because some substrates may enrich the environment and support important behaviors of the birds (Gunnarson et al., 2000), as well as determine chickens' skin condition (Mendes et al., 2011).Thus, providing a good litter would be an effective way to increase broiler activity (Shields et al., 2005) and to reduce locomotion problems (Almeida Paz et al., 2010).A study carried out by Newberry (1999) assessed the effect of increasing the environment complexity on the use of broiler pen space, determining that the presence of specific materials stimulates exploratory behavior, with the consequent beneficial effects on the locomotive system of these birds.One type of exercise in birds is dust bathing, but there are other behaviors, such as pecking and scratching which are also said to be beneficial.In addition, this environmental enrichment can be devised in several ways; for example Arnould et al. (2004) assessed the effect of enriching experimental pens with sand trays, finding that they were attractive for chickens.
Laying hens have shown a strong preference for certain substrates such as sand to practice those cited activities (Nicolet al., 2001), while broilers clearly prefer this material to perform most of their dust baths when four different substrates are made available (Shields et al., 2004).In addition, it seems that the availability of sand enhances chickens' foraging activities (Arnould et al., 2004).Most of these studies have focused on certain behaviors such as dust bathing, pecking or scratching, but only a few of them studied other behaviors on these substrates (e.g.Lindner & Hoy, 1997;Shields et al., 2005).
Several of the cited studies have been carried out through preference or choice tests (Shields et al., 2004;Sanotra et al.,1995).These two terms are not the same, according to Kirkden and Pajor (2006): preference denotes a difference between the strength of motivation to obtain or avoid something, while choice describes characteristics of an animal (e.g., an animal prefers bananas to oranges).In recent years, both of these tests have been used with poultry to develop preference studies with very different objectives, such as studying color in pecking devices (Jones & Carmichael, 1998), light sources (Moinard & Sherwin, 1999;Kristensen et al., 2007), type of cages (Elston et al., 2000) or ammonia concentrations (Kristensen & Wathes, 2000).Therefore, the first aim of the present research was to assess the selection by broiler chickens of different types of materials widely used as litter in the poultry industry (sand, wood shavings, straw and rice hulls) using a choice test.The second objective of this study was to determine the predominant behaviors on each of these materials.

Animals and housing
In this study, 40 one-day-old male Rossâ broiler chicks (Aviagen, Alabama, USA) were obtained from a commercial hatchery.Birds were housed in the same home pen and marked with rings on their right legs in groups of eight birds.Four ring colors were used, one for each bird group, and a group of eight birds had no mark, so these could be used to replace any of the marked broilers, if needed.This replacement was never necessary, so 32 chickens were finally used in the choice test experiment.The home pen measured 2x2 m and contained wood shavings, sand, rice hulls and straw.These substrates were located in four different sections of the pens, each section measuring 1 m 2 (see Figure 1) with a substrate depth of 15 cm.The chickens could move freely around the 4 m 2 pen, so they had permanent access to the four materials.In the center of the pen a drinker and a feeder were located, so that broilers could drink or eat from any part of the pen and, in consequence, while standing on any of the four substrates.Feed and water were provided ad libitum and for the first three days, chickens had 24 hours light, which was then gradually decreased to a 16L:8D schedule.No birds died during the experiment.

Choice test
Choice test was carried out when birds were between 14 and 42 days old (for four weeks), three days per week according to the following protocol.One of the marked groups was transferred from the home pen to the choice test pen, which was located in the same room.This pen was exactly the same as the home one, as shown in Figure 1, and the birds could freely access the four studied substrates, a feeder and a drinker.
Once the eight chickens were in the choice test pen, the observer sat quietly about 2 meters away from the front of the pen and the chicks were given 10 minutes Gómez EA, Lainez M, Torres AG

Behavior of Broiler Chickens in Four Different
Substrates: a Choice Test to habituate to the observer's presence and to the new pen.Then, behavioral and choice observations started.This protocol was adapted from Shields et al. (2005).
Every five minutes for one hour (12 observations per hour), a scan sampling was carried out, and the position of each bird and the performed behavior was recorded.After the first hour, the group was removed, the test pen was cleaned and another group was transferred from the home pen to the choice test one, and the same protocol was followed.The procedure was identical for the four groups each day the choice test was done.A Latin square was used to determine the order in which the groups had to be transferred to the choice test pen.
The observed behaviors were classified into the following categories: resting, preening, walking, eating, drinking, dust bathing, fighting, standing, pecking, scratching, and flapping.This ethogram is explained in Table 1.The experimental design was approved by the Ethical and Animal Welfare Committee of the Valencian Institute of Agricultural Research.

Statistical analyses
Original data were converted into frequencies, so that in each observation (every five minutes) there was only one data item including the eight animals, expressing the relative percentage of animals that were performing one activity ona specific substrate.The experimental unit was the group, with bird nested within group.
General distribution of frequencies was analyzed using a chi-squared test (procedure FREQ of SAS â System, SAS Institute 2009).Subsequently, in order to identify the activities that were preferably performed by the broiler chickens on each of the four studied substrates, a factorial discriminant analysis, using stepwise option, was run with typified variables (STEPDISC procedure from SAS System, SAS Institute, 2009).
On the other hand, in order to assess the effect of each substrate, bird age and time of day on the different observed behaviors, a logistic regression model was performed (Agresti, 1990), using procedure GENMOD of SAS System (SAS Institute, 2009), assuming a binomial distribution and using log it as the link function.In addition, the GLIMMIX procedure by SAS System (SAS Institute, 2009) produced the estimates of the average log its on the scale of the data.
The evaluated effects were substrate type , the time of day (because the test was carried out during the entire morning each day, and it was possible that the time of day had some influence) and bird age, so repeated measures were included.For this effect, it was decided to divide the test into two different ages: from the beginning (14 days) to 21 days old (the middle of the rearing period) and from day 21 to the end (day 42).The interaction between substrate and age was also evaluated.As a result, the equation (1) shows the logistic regression model: Where: L ijkl =linear logistic model; logit(P ijk ): the probability of a categorical response (P p =probability of "positive activity" response; 1-P p =probability of "absence of activity" response); b 0 =intercept; b 1 , b 2 , b 3 = coefficients estimated for the logistic regression models; T i =Time of the day (from 1 to 240 minutes);

Behavior of Broiler Chickens in Four Different Substrates: a Choice Test
A j =effect of age, (age ≤21day, considered as age1; or age>21day, considered as age 2); S k =effect of litter material (k=1, straw; k=2, wood shavings; k=3, rice hulls; sand, used as reference); e ijkl = residual error.Likelihood ratio tests were performed among nested models for computing likelihood-based fit statistics.

RESULTS
Figure 2 shows the general distribution of the frequencies of the presence of the birds on each of the four studied substrates.The frequency of the birds on sand was the highest, followed by wood shavings (c 2 = 1867.91,p<0.0001).The discriminant analysis maximizes the differences between the four floor substrates.It uses linear combinations of the frequencies of the performed activities by the birds on each bedding material.The selected functions using the "stepwise" option in factorial discriminant analysis are shown in Table 2 and graphic results of this analysis are observed in Figure 3.In Table 2, only resting, preening, eating, drinking, dust bathing, pecking and scratching are displayed, given their statistically significance (p<0.05) in the model, while walking, sitting, standing or flapping did not.This means that they do not help to explain the differences between the bedding materials.
The first discriminant function divides substrates where preening, eating and dust bathing (positive values) were preferentially performed, from those which presented opposite values, mainly resting and drinking.On the contrary, the second discriminant function differentiates the substrates where preening, drinking and dust bathing performed from those where birds demonstrated pecking and scratching.Combining these two discriminant functions, Figure 3 was obtained, where bird groups are distinguished among substrate.This figure shows that wood shavings and straw were very similar because they clustered, while rice hulls and sand were clearly differed from both and even between each other.Therefore, three groups were clearly differentiated and it was expected that the behaviors performed in each of those groups were different.
Relative to the application of the logistic regression model, Table 3 shows the significance of each effect evaluated in the logistic model, and it may be observed that the proposed logistic model fitted all of the activities, except for flapping.It is also observed that depending on the behavior, the significance of the effects differs.The effect of substrate was statistically significant for all behaviors, Gómez EA, Lainez M, Torres AG

Behavior of Broiler Chickens in Four Different
Substrates: a Choice Test except for fighting and standing, whereas the effects of the time, bird age and their interaction with substrate effects were more dependent on each behavior.
Once the logistic regression model was fitted, coefficients were obtained for each behavior (Table 4).These coefficients reveal the real influence of time, age and substrate on each behavior; the higher the value of the b coefficient, the greater the frequency of that behavior in relation to sand and age >21 days, which were used as references in the logistic model.Similarly, the lower the b coefficient, the lower the frequency of a behavior.For instance, compared with sand, scratching was performed more in any other substrate, although the highest frequency took place on rice hulls, as b was the highest estimated logistic regression coefficient.All the values shown in Table 4 and those related to the interaction between age and substrate were used to construct Figure 4 and Figure 5 (drinking and fighting).These figures illustrate more clearly than Table 4 that certain behaviors were preferably performed on some substrates.For example, dust bathing was mainly performed on sand, and pecking and scratching were mostly developed on rice hulls.Likewise, the effect of age on each behavior and substrate is clearly shown.
The most remarkable results were obtained for resting, dust bathing, pecking, scratching and preening.Resting was performed mainly on wood shavings and straw, which were preferred over sand and rice hulls (although when birds were older, they also tended to rest on sand, as shown in Figure 4).On the other hand, age tended to decrease the probability of this behavior, although an interaction with substrate was present, so this pattern changed to sand.In contrast, sand was strongly preferred for dust bathing, while rice hulls were chosen for pecking and scratching.In addition, preening was performed in all substrates, although wood shavings and sand were those in which preening was most frequent.Age also had a notable effect on this behavior and it was coincident with dust bathing tendency with age, and occurred mainly on wood shavings and sand.Furthermore, the effect of, time of the day on certain behaviors was statistically significant, so as the morning (and hence the test) progressed, results varied.This is shown in Figure 5, where it can be observed that the tendency to fight decreased over the morning.Fighting was a priori of little importance, but we detected that it was only observed in the first moments of the morning and then the probability of the display of this behavior decreased (b 1 =-0.087).In contrast, drinking followed a very different pattern, increasing as the morning advanced.In both cases, this pattern was found regardless of the age of the birds and the substrate on which the behavior was performed.

DISCUSSION
Sand was strongly selected over the other three types of litter in terms of the general results, but this pattern changed when individual behaviors were assessed.Although the current substrate can determine Gómez EA, Lainez M, Torres AG

Behavior of Broiler Chickens in Four Different
Substrates: a Choice Test the behavior or the chosen substrate in laying hens (Nicolet al., 2001), that was not possible in our study because broilers were able to access the four bedding materials in the home pen, so in this study sand was clearly chosen.
When behaviors were separately evaluated, sand was mainly chosen for dust bathing.This preference for sand was previously reported by Shields et al. (2004) and Sanotra et al. (1995), and it is quite clear that if the birds have the possibility, they perform these baths on sand.In addition, it was observed that wood shavings were more frequently chosen than rice hulls or straw to perform dust bathing.Simultaneously, preening was mainly performed on sand when birds were older than 21 days and on wood shavings when they were younger than 21 days.These results may indicate that it is a kind of behavior following or prior to the bath, as it coincides with dust bathing results, although the effect of the substrate itself is less clear.
Broilers selected rice hulls for pecking and scratching, so they did not peck more frequently on the same substrate they preferred for dust bathing, as suggested by Shields et al. (2005) and Arnould et al. (2004).Therefore, our results are not consistent with those that suggest that the behaviors of pecking and scratching are previous to dust bathing.However, pecking frequency may be related to dust bathing, because it increased when birds were older than 21 days, which coincided with the higher frequency of dust bathing.However, a possible explanation for our findings is that pecking and scratching are exploratory behaviors, and therefore certain substrates that are friable, such as rice hulls, may attract the birds to peck or scratch.On the other hand, this material may not be adequate or stimulating for dust bathing because it maybe too abrasive.The present results cannot be compared with other studies, such as that of Sanotra et al. (1995) because they did not use rice hulls as substrate; however, they found that scratching was very frequent in straw, and that scratching and pecking preferences were different from those for dust bathing.Gunnarsson et al. (2000) also reported that straw was preferred for pecking and scratching, but not for dust bathing, but did not test rice hulls either.
The hypothesis that rice hulls may be abrasive is supported by there sting frequency results.Broilers selected to rest on, in decreasing order wood shavings, straw, sand, and rice hulls, indicating that wood shavings and straw are more comfortable than rice hulls.Furthermore, it must be remembered the strong tendency of chickens to peck and scratch on rice hulls, and therefore, other behaviors may be much less likely to occur.Relative to resting on wood shavings and sand, our results were opposite to the findings of Shields et al. (2005), who observed that resting as well as active behaviors were performed more on sand; however, they did not test sand as substrate.Lindner et al. (1997) also observed that resting frequency was significantly higher on wood chips than on straw.
Walking and standing were studied as they indicated general activity, but as seen in Figure 4, their frequencies were very low (maximum of 1.2% and 0.8%,respectively), and therefore, no conclusions can be drawn from these data.
Taking all these findings into account, more than one litter substrate should be available to broilers on the farms to allow them to perform some behaviors, which may be compromised in current conditions.
Age influenced some behaviors (Table 3), differently from the findings of Shields et al. (2004).The individual frequency of the evaluated behaviors decreased with age(b 3 in Table 4), except eating, which increased as birds aged.However, as shown in Figure 4, the interaction between age and substrate had a different effect, and the frequency of certain active behaviors(such as dust bathing, pecking and scratching), tended to increase with age when birds were on determined substrates, which is just the opposite to that found by Shields et al. (2005).A possible hypothesis to explain our results is the design of the choice test.Broilers were transferred to the choice test pen every experimental day, but they lived in the home pen, which litter became gradually dirty.On the other hand, bedding materials in the choice pen they were clean, unused and attractive, and therefore the birds preferred to explore and use them instead of eating or resting on them.This exploratory behavior may also explain the lack of observations in which the birds were just walking or standing, as mentioned above.
On the other hand, the effect of time of day was statistically significant on certain behaviors (Table 3).Fighting decreased as the morning advanced.Although we hypothesized that this may be due to general activity reduction over the morning, the remaining active behaviors did not confirm this hypothesis, and we did not find any literature regarding the daily pattern of these behaviors.The strong increase in drinking behavior shown in Figure 5 during the morning maybe due to increase in room temperature, although these data were not available and therefore further research is necessary to identify the causes of these findings.

Figure 1 -
Figure 1 -Distribution of the home pen and the identical test pen

Figure 2 -
Figure 2 -Distribution of frequencies at which birds chose each studied substrate

Figure 3 -
Figure 3 -Scattered plot per substrate of the relative frequencies of activities in two discriminant dimensions

Figure 5 -
Figure 5 -Frequencies of drinking and fighting, which are influenced by time of day

Table 1 -
Developed ethogram during the choice test

Table 2 -
Coefficients of Factorial Discriminant Analysis for litter variable

Table 3 -
Results of the application of Logistic RegressionModel to the assessment of the effect of substrate, age and time of day on different behaviors of broiler chickens

Table 4 -
Estimated coefficients throughout the Logistic Regression Model for the different behaviors