The problem of demonstrating that animals are consciously aware of their environment or themselves in a similar manner to ourselves has proved to be a relatively intractable one since we have mainly been constrained to rely upon observation of an animal’s behaviour as an index of its awareness. It can usually be argued that even complex behaviours need not necessarily involve an animal being aware of either environmental cues or its response to them. Indeed, simple organisms or computers can perform complex responses to environmental cues when they are clearly not aware of them in the sense that we are ourselves. What can particularly distinguish conscious awareness from simple stimulus-response guided behaviour is the ability of an animal to experience some form of emotional response to an environmental cue or to its own actions. In this review I will summarize some of our experiments on sheep which show that the way their brains process visual and olfactory social and non-social cues from the environment is integrally related to and organized by the emotional significance of what is perceived. I will also argue that the way their brains process complex social stimuli from the environment in order to recognize individuals is very similar to ourselves although these processes are very strongly influenced by changes in their motivational state. From these observations I will maintain that sheep are almost certainly capable of conscious awareness of their environment although perhaps this awareness is mainly restricted to present events and needs with a limited capacity to reflect on the past or future.

Measuring motivation has recently become a key issue in animal welfare, yet it can be difficult to implement in practice and even more difficult to apply validly to the specific animals whose welfare it is hoped to improve. Ethologists have modelled motivation in a number of ways. Here, we review these models (along with consumer demand approaches) to identify some of the factors that need to be controlled to conduct experiments with maximum internal and external validity. They indicate that to conduct experiments that make valid assessments of animals’ priorities, bouts of behaviour should not be curtailed, measurements should not be restricted to only one period or context and subjects should be kept in closed economies; time spent with resource should not be used as the only measure of consumption, as rate can vary with motivation and if demand curves are desired, the cost paid and amount of opportunity ‘bought’ must co-vary. Having avoided these pitfalls, further factors must be taken into account to ensure external validity. Animals’ priorities are affected by many aspects of their internal state and external environment, including the presence of eliciting stimuli, the number of behavioural opportunities available and the size of their time and energy budgets. A well fed animal in an enriched enclosure with excess energy but only limited time to allocate to many different activities would thus be likely to have quite different priorities from an under-fed animal with excess time available, housed in a barren environment. Hence studies of the former could not validly be applied to improve the welfare of the latter.

Current methodologies for measuring choices are the product of early psychological research into behavioural mechanisms, particidarly learning. In applied behaviour studies, much of the emphasis has been on the use of preference tests to assess animals’ motivation, and hence proximate need for, various forms of stimulation. Preference tests are based on a modular view of animal behaviour, where animals are presented with physical compartments each containing different resources such as food or social partners. An obvious limitation of the approach is that simple preference tests may do little justice to the complexity of animals’ behavioural organization and environmental preferences. Variability of choice behaviour also poses a considerable problem, as it may prevent us understanding and interpreting short-term choices and consequently describing the proximate needs of the animal. We also believe that the approach of applying consumer economics to quantify the value of resources to animals is increasingly moving preference testing away from measuring proximate to measuring ultimate need. This shift in emphasis does not appear to have been recognized. The risk is that the approach will only identify basic needs (e.g. related to growth and reproduction) but be insensitive to the proximate behavioural needs that it set out to quantify. In general, the focus of welfare-related preference testing should shift from the measurement and description of preferences to the more strategic task of understanding the rules governing short-term choices. We will be unable to measure proximate need unless we can develop a theoretical framework better able to interpret short-term behavioural choices.

Human-animal interactions are a common feature of modern intensive farming systems and these interactions may have marked consequences on animal productivity and welfare. Research in agriculture has shown interrelationships between the stockperson’s attitudes and behaviour and the behaviour, productivity and welfare of farm animals and the following model of human-animal interactions in agriculture has been proposed. Because a stockperson’s behaviour towards animals is largely under volitional control, this behaviour is strongly influenced by the attitudes that the stockperson holds about the animals. These attitudes and consequent behaviours predominantly affect the animal’s fear of humans which, in turn, affects the animal’s performance and welfare. The mechanism whereby fear affects performance and welfare appears to be through a chronic stress response. The risk to welfare also arises if the stockperson’s attitude and behaviour towards the animals are negative because the stockperson’s commitment to the surveillance of, and the attendance to, welfare issues is likely to be highly questionable.

Recent research in the pig industry has shown that a training programme targeting the key attitudes and behaviour of stockpersons produced improvements in stockperson attitudes and behaviour and animal behaviour and productivity. These results indicate the potential to improve animal productivity and welfare by targeting the stockperson’s attitudes and behaviour for improvement. Techniques which may be useful in this regard include staff selection and training procedures addressing these important human attributes.

To assess farm animal welfare we need to understand how animals make choices and how these choices relate to preference strength. Studies of environmental choice can be categorized by the method used to investigate them, or by the underlying basis on which the animal is choosing. Choices made between resources that vary along a single dimension should meet certain criteria e.g. those of transitivity. Choices made between resources that vary along more than one dimension may or may not meet these criteria, depending how the animal evaluates each option. Understanding how farm animals choose will allow the results of individual experiments to be applied in a wider context. It is also important to know how preferences are formed during development. Evidence suggests that preferences for nests and pecking substrates in hens may be influenced by prefunctional experience. Experimental data from studies of environmental choice may enable us either to provide important resources in commercial systems, or to provide facilities for animals to continue to make their own decisions.

From an animal production and welfare perspective, an important social choice made by farm animals is whether and how vigorously to fight others. The choice to fight (i.e. simultaneous escalated aggression by both contestants) may result in severe injury when unfamiliar animals first meet or compete for highly valued resources. Game theory models of aggressive interactions predict that animals have evolved to stop or avoid fighting when they assess their chances of winning a contest to be poor. Assessment may occur during or before a fight. If it occurs before a fight, reliable cues of fighting ability must exist. If farm animals can establish relative social status by assessment prior to fighting, it may be possible to construct social groups which contain individuals of differing abilities such that most disputes are resolved through assessment and levels of damaging aggression are kept low. Social status appears to be established by assessment rather than fighting in free-ranging deer and sheep when asymmetries exist in reliable cues of their fighting abilities such as roaring rate and horn size. However, there have been few detailed studies of other farmed species and findings are equivocal. Work on young pigs suggests that, in pair-wise encounters, individuals are unable to assess weight-related asymmetries in their abilities without fighting. However, recent studies of groups of pigs suggest that some form of assessment prior to fighting may occur. Individuals were classified as high (H) or low (L) aggressive on the basis of their behaviour in an attack latency test. When litters of H pigs were mixed with litters of L pigs, significantly fewer pairs of unfamiliar pigs fought than when newly mixed groups were made up of litters of H pigs only, or litters of L pigs only. Thus, fighting was least frequent when there was a marked asymmetry in the aggressiveness of unfamiliar individuals. Another study raised the possibility that H and L pigs may be following alternative strategies which, under certain circumstances, are similarly beneficial in welfare and production terms. Further work is required to substantiate these findings and to determine whether aggressiveness is a reliable cue of fighting ability and, if so, how it is manifest and assessed.

Irrespective of whether farm animals are given access to a single homogeneous food or to two or more heterogeneous foods as a choice, their feeding behaviour raises one of two questions (i) how much to eat, and (ii) what to eat? Despite frequent comment to the contrary, their feeding behaviour appears to be goal-orientated rather than random or purposeless. I therefore consider first the goals of farm animals in relation to their feeding behaviour. In general it is accepted that the overall biological framework in which all animals are trying to maximize fitness’ also applies to farm animals. However, some modification is called for in order to account for those situations in which intensive genetic selection has led to relatively ‘unfit’ reproducing animals and for those cases where animals are given access to foods which have not figured in their evolution

On the basis that feeding behaviour is goal orientated, I then consider whether farm animals achieve their goals by monitoring their behaviour in the short or longer term. The conclusion drawn is that while short-term feeding behaviour may be a device to exploit the feeding environment effectively, it is largely unrelated to short-term fluctuations in an animal’s internal state. By contrast, longer-term feeding behaviour is very closely related to longer-term change in internal state, implying the maintenance of close control over feeding behaviour in terms of food intake and diet selection. All animals, including farm animals, are considered to be creatures of habit which maintain habitual feeding behaviour until a change is provoked by a significant alteration in their internal state. Such an alteration requires to be of significantly large magnitude and to be unlike the usual short-term, systematic fluctuations which occur over a day in the profiles of metabolites or hormones. Based on this premise, I contend that the mechanisms by which these disturbances are perceived by the animal will be general rather than specific. The notion that animals can fully achieve their goals by monitoring their feeding behaviour is obviously applicable in situations where they are given appropriate nutritional choices. Where animals are given inadequate or inappropriate choices, as is predominantly the case with farm animals, their feeding behaviour is designed to bring them as close as possible to their goals. Finally I consider the relevance of nutritional choices to farm animals by addressing the possibility of exploiting the goal orientation of feeding behaviour. I conclude that greater recognition of the goal-orientated nature of farm animals’ feeding behaviour can bring benefits in three areas: (i) improved biological understanding of animals’ goals; (ii) improved animal welfare; and (iii) improved animal performance.

The means by which choice can be engineered into farm animal environments are described, focusing upon intensive systems of husbandry where choices are the most limited. Choice is assumed to be both necessary and desirable.

The principal components of choice are: (i) environmental choice; (ii) nutritional choice; and (iii) social choice. The management of herds rather than individuals and the persistent trend towards increased intensification generally reduces the opportunities for animals to exercise their environmental and social choices. However, extensive systems can also restrict choice and provide an environment that may, in some cases, be equally barren.

Several common problems may make the provision of choice difficult, including variability and precision of individual choices and the conflicts between the interests of individual animals (as distinct from groups) and farmers themselves.

The availability of choice can be improved by modifying existing systems or developing new systems of livestock husbandry. The choice of physical environments may be made by self selection by an individual animal in a heterogeneous environment; active animal control of an environmental modifier, or automatic control with continual monitoring of physiological and behavioural states. The enrichment of barren environments could also embrace artificial stimulation of the senses. Improving the choice of space for its many social and other uses is likely to be difficult given the financial constraints of livestock husbandry. There are few technical difficulties to improving nutritional choices available to livestock, e.g. automated feeding systems which provide multiple rations.

The concept of palatability occurs frequently in studies of voluntary food intake and diet selection (see for example the numerous references in Forbes, 1986). The purpose of this paper is to suggest that the concept is of limited value. Four major areas are considered. The first is that of definition, as the term is often used in an ambiguous manner. Secondly, the use of palatability as a food descriptor is challenged since it is essentially an animal rather than a food characteristic. Thirdly, the usefulness of the concept is challenged in the context of its evolutionary and ecological significance. Finally, it is contended that the response of the animal is not necessarily a good measure either of the stimulus received or of the resulting internal state.

Locomotion is recognized as an important aspect of behaviour and knowledge of the locomotion of broiler chickens is important for their health and welfare (Lewis and Hurnik, 1989). Würbel (1995) suggested that certain behaviours can give an indication of poor welfare and that preference tests may be used to ‘fine tune’ a housing system.

Voluntary automatic milking is proposed as a system whereby dairy cows can choose to be milked when they want. The aim is to develop a system that can milk and perform all the associated tasks without requiring routine human intervention. (Details of the design and operation can be found in Street et al., 1992.) The system may not be viable unless the cows visit at an appropriate frequency, high enough to generate the 0·10 to 0·15 proportional increase of milk yield from cows milked three or more times per day (e.g. Knight and Wilde, 1993) but not so high as to result in some cows over-using the system. Understanding why cows may want to visit the system is therefore important. The two most important reasons why a cow may attend are likely to be motivation to be milked and motivation to eat. Motivation to be milked may change as lactation progresses. Late lactation cows have been shown to enter a milking parlour later than high yielding cows in one experiment (Rathore, 1982) but not in another (Winter, 1993). Rathore (1982) suggested that motivation to be milked may be generated by the discomfort of a large and distended udder. Motivation to be milked may also be linked to some inherent desire of the cow to suckle and subsequently wean her calf. This may be independent of the amount of milk in her udder and decline during lactation. Therefore motivation to be milked could be generated either from the discomfort of a large and distended udder or by some psychological desire by the cow to suckle a calf generalized to a milking machine. Either way, the strength of motivation to be milked may have implications for how an automatic milking system (AMS) is designed. For example, if cows were highly motivated to be milked and attended at an appropriate frequency, there would be no requirement to provide additional incentives, such as food, to attract cows into the AMS. In addition, if cows choose to be milked more frequently than they are milked in conventional systems (generally twice per day), then this may be a method by which choices can be engineered into an animal’s environment.

Recent grazing preference studies show that sheep offered a choice between large adjacent monocultures of perennial ryegrass and white clover choose a diet that is a mixture of the two species (ca. 0·7 white clover) (Parsons et al., 1994). One possible hypothesis to explain this is that animals need to sample different areas to know what is located there or the different plant species to know what eating each species represents and because of this include apparently less desirable foods (here ryegrass) in the diet (see Parsons et al., 1994). Edwards et al., (1996) showed that sheep can use spatial memory and so may not need to sample on an area basis. In this study, we tested directly whether sheep could distinguish between ryegrass and white clover without grazing them and whether they can form associations between the appearance of the vegetation and the food reward it represents.

Plant species or vegetation characteristics (e.g. sward height) are often distributed in patches within a background of continuous vegetation. Grazing animals exploit this spatial heterogeneity by concentrating their foraging in patches that are of preferred species or in patches that offer high rates of intake (Bazely, 1988). However, little is known of the mechanisms or individual behaviours that animals use to accomplish this preferential patch use. One hypothesis proposed to account for the preferential patch use is that animals remember information about the spatial distribution of the patches they encounter and use this information on subsequent foraging bouts to increase the rate at which they encounter preferred food patches (Bell, 1991). In this study, we tested whether sheep could remember the spatial location of patches in simple and complex environments or whether they needed continually to sample to know what was located in each area.

During the process of transportation sheep can experience long periods of food deprivation during the journey, at market and in lairage. New European Union proposals will allow sheep to be transported for up to 14 h without food. However, it is not clear how such periods without food affect the welfare of sheep.

Dietary choices of herbivores on rangeland complex vegetation are not easy to assess and give more or less biased results despite dedicated work (Holecheck et al., 1982; McInnis et al., 1983). Direct methods, such as oesophageal fistulas and direct observations can only be used in a limited range of conditions. Their negative effect on animal welfare emphasizes the need for non-injurious and nonstressful methods.

Domestic sheep have a strong social tendency. However, the nature of social relationships varies with season, breed, and size of enclosure. Ewes occupy a smaller home-range, with larger subgroup sizes, in the winter compared with the summer months (Lawrence and Wood-Gush, 1988a). Domestication has apparently led to a reduction in nearest neighbour (NN) distances from feral (Soay) sheep, who do not have strong flocking behaviour, to Merinos with NN distances of less than 2 m (Arnold, 1985). The social choices made by ewes affect their foraging behaviour and may have significant effects on production (Lawrence and Wood-Gush, 1988b). The aim of this study was to examine the social choices of a hill and a lowland breed of ewe when grazing two different environments in spring and summer of 1995 and 1996. The influence of lamb genotype on the behaviour of the ewes was also investigated.

Sheep in the uplands of the United Kingdom face a series of choices as to where to graze. These choices have been shown to be influenced by the spatial distribution of nutrient supply from herbage (Duncan et al., 1994). Micro-climatic variation may also determine where animals forage, particularly in topographically complex environments found in upland areas. Wind and rain have marked effects on heat loss and hence on lower critical temperature (LCT) in sheep (Blaxter, 1977). Mount and Brown (1982) demonstrated that there were occasions during a year when sheep were below their LCT. This study assumed that data gathered from a standard meteorological station were applicable to conditions actually experienced by grazing animals and took no account of the ability of herbivores to select a more favourable micro-climate under extreme conditions (Mount and Brown, 1982). The experiment reported here was carried out to examine the extent to which microclimate influenced the distribution of grazing by sheep under windy conditions. Micro-climate was manipulated using artificial shelters.

To develop sustainable grazing systems, an understanding of the complex interactions between competing plant species and grazing herbivores is needed. An understanding of dietary preference is a prerequisite to predicting how much and from which plant species animals select. This work initially concentrated on dietary preferences of sheep grazing perennial ryegrass and white clover swards and was then extended to other animal species to test the generality of our findings.

The existing legislation in England and Wales (Ministry of Agriculture, Fisheries and Food, 1995) prohibits slaughter (sticking) of red meat animals within sight of others. Although this prohibition was originally designed to protect the welfare of animals, trying to get stunned animals out of sight of others can often result in signs of recovery being shown prior to, during and following sticking in commercial situations (Anil and McKinstry, 1993). The question of whether slaughter by exsanguination (sticking), or killing within sight, causes stress in other animals is a controversial issue. In regard to witnessing slaughter we have already attempted to address this question. This work produced no evidence to suggest that sheep (Anil et al., 1996) or pigs (Anil et al., 1997) were distressed. The effects of killing by mechanical means or shooting animals on witnessing conspecifics have also been studied in mice, hens and farmed deer by Bracke (1993) who concluded that no evidence of distress was evident. However, it is still widely believed that slaughter animals are sensitive to the presence of blood at abattoirs. This investigation was designed to test whether sheep would show any changes in their behaviour when exposed to the sight and smell of conspecific blood.