The idea of optimisation in animals: uses and dangers

Diet selection in pigs: choices made by growing pigs when given foods differing in nutrient density

Two experiments were conducted to corroborate or refute the theory that animals will choose a food that will allow them to use it with maximum efficiency. Pigs have been shown to utilize foods of high nutrient density more efficiently than those of low density, so the choices made by pigs when offered such foods could be used to test the above optimization theory. In experiment 1, 48 Large White × Landrace gilts were used, for an 8-week period starting at 22 kg live weight, while in experiment 2, 48 boars of the same cross but of a genetically improved strain were used from 24 to 60 kg live weight. In both experiments use was made of high nutrient density summit foods which were used alone, or diluted in the ratio 80 summit: 20 milled sunflower husk to provide the low density foods. In experiment 1, the high density diet (HI) contained 7·5 g lysine per kg and 13·20 MJ digestible energy (DE) per kg, whereas in experiment 2 two summit foods were formulated, the first diet (H2) was offered for 3 weeks from 24 kg live weight and the second (H3) followed until 60 kg live weight. Foods H2 and H3 contained 11·0 and 8·40 g lysine per kg respectively and 15·0 and 14·0 MJ DE per kg, respectively. Both experiments made use of a high (H1 and H2, respectively) and a low nutrient density (L1 and L2, respectively) control treatment in which pigs were given ad libitum access to H1 and H2/H3, and L1 and L2/L3 in experiments 1 and 2 respectively (no. =4). In addition, a medium density treatment (Ml) consisting of a 50: 50 mixture of H1 and L1 (no. = 4) was given in experiment 1. Two choice-feeding treatments where used in both experiments, the first in which H1 and H2IH3 were placed in the left bin (CL1 (no. =18) and CL2 (no. = 20), respectively) and the appropriate dilution diet in the right bin, and the second in which H1 and H2/H3 were placed in the right bin (CR1 (no. = 18) and CR2 (no = 20)). There were no differences in average daily growth rates between treatments within experiments but there were significant differences (P < 0·05) in food intakes and efficiency of food utilization (FCE) between treatments. The highest intakes and lowest FCE were obtained on the L1 and L2 treatments while the lowest intakes were recorded on the choice-feeding treatments. There were no significant differences in FCE neither between H1, CL1 and CR1 nor between H2, CL2 and CR2. Only in experiment 1 were there significant differences (P < 0·05) between choice-feeding treatments on the basis of the position of the food bin but there was no preference for a particular position. The results indicated that pigs were able to differentiate successfully between two foods on the basis of their nutrient density, that bin position was not used as a cue in the choice made, that a small amount of the ‘unwanted’ food was consumed throughout the experiment and that the diet selected maximized FCE.

Tests of two theories of food intake using growing pigs 2. The effect of a period of reduced growth rate on the subsequent intake of foods of differing bulk content

The effect of a period of feeding on a high bulk food, upon the subsequent intake of foods of differing bulk content, was investigated in two experiments of the same design. The intention was to provide a severe test of the two current conceptual frameworks available for the prediction and understanding of food intake. In each experiment 40 male Manor Meishan pigs were randomly allocated to one of four treatment groups at weaning. Each experiment was split into two periods, P1 (12 to 18 kg) and P2 (18 to 32 kg). The treatments, all with ad libitum feeding, were: a control food (C) given throughout (treatment CC); a medium bulk food (M) given throughout (treatment MM); a high bulk food (H) given in P1 and then C in P2 (treatment HC); H given in P1 and M in P2 (treatment HM). C was based on micronized wheat with 13·4 MJ digestible energy and 243 g crude protein per kg fresh food. In experiment 1 M contained 350 g/kg and H 560 g/kg of unmolassed sugar-beet pulp and in experiment 2 M contained 500 g/kg and H 700 g/kg of unmolassed sugar-beet pulp. Framework 1 predicted that food intake on the medium bulk food (M) would not be increased, whereas framework 2 predicted that intake on M would be increased after a period of feeding on H, compared with when M was offered continuously.

In P1, both food intake (P < 0·01) and growth (P < 0·001) were severely limited on H compared with C. In experiment 1 growth was limited on M compared with C during the first 7 days of P1 (P < 0·01) only. In experiment 2 intake (P < 0·001) and growth (P < 0·001) on M were limited throughout P1, compared with C but not thereafter. Therefore, in neither experiment did M cause a lower growth rate than C from 18 to 32 kg. In experiment 1 there was full adaptation to M after about 10 days from 12 kg. In experiment 2 adaptation was complete by the end of the first 7 days from 18 kg.

In P2, food intake (P < 0·001) and live-weight gain (P < 0·05 and P < 0·001 in experiments 1 and 2, respectively) were increased on HC compared with CC. By the last 7 days of P2 intake was still higher (P < 0·01) but growth rate was no longer different to CC. Intake and gain were increased in P2 on HM compared with MM but, in general, these differences were small and not significant. In the first 7 days of P2, in experiment 1 pigs on HM had higher intakes (P < 0·001) and gains (P < 0·05) than those on MM, but in experiment 2 only intake was higher (P < 0·01) with no difference in gain. By the last 7 days of P2 there was no difference in either intake or gain between these two groups in either experiment. Pigs on HC increased intake by more than those on HM. There was, therefore, a significant interaction for food intake (P < 0·05, in experiment 1 and P < 0·001, in experiment 2) between prior and present food.

The unexpected failure of either M food to limit growth throughout the experimental period meant that the results of these experiments could not be used as a strong test to reject either one of the frameworks. However, the ability of the pigs to compensate on M was less than that on C. The data provide some evidence that under conditions of compensation foods such as M may be limiting. This is in closer agreement with the framework that predicted that consumption of a limiting food will not increase after a period of feeding on a high bulk food (framework 1).

The effect of choice feeding complete diets on the performance of weaned pigs

Post-weaning growth rate in pigs is frequently poor and variable. Choice feeding may offer the opportunity to rectify this. In experiment 1, 24 mixed sex groups of 16 pigs weaned with an average weight of 6·8 kg were blocked on weaning weight and assigned at random to the following treatments: (A) starter diet (18·3 g/kg lysine and 16·6 MJ digestible energy (DE) per kg) for 11 days followed by link diet (15·0 g/kg lysine and 15·3 MJ DE per kg) to 27 days; (B) starter diet and link diet offered in a free choice to 27 days; and (C) starter diet and weaner diet (13·0 g/kg lysine and 14·3 MJ DE per kg) offered in a free choice to 27 days. In experiment 2, 66 pigs were weaned at 6·3 kg, blocked as individuals on sex and weight and were randomly assigned to treatments as in experiment 1. In experiment 3, 24 single sex groups of 16 pigs were formed from pigs weaned at 7·8 kg. The groups were blocked on the basis of weaning weight and randomly assigned to the following treatments: (A) starter diet for 11 days followed by link diet to 26 days, (B) starter diet and link diet offered in a free choice to 26 days and (C) starter diet and link diet offered in a free choice to 26 days with feeder position rotated twice weekly. In experiments 1 and 3 pigs were offered a common weaner diet (14·1 g/kg lysine and 14·4 MJ DE per kg) following the experimental period. In experiment 1, daily gain was 406, 410 and 397 g/day (s.e. 6·6; P > 0·05) and food coversion efficiency (FCE) was 1·24, 1·21 and 1·27 g/g (s.e. 0·01; P < 0·01) during the period from day 0 to 27 for treatments A, B and C, respectively. The proportion of the diet selected as starter diet was 0·20, 0·50 and 0·47 (s.e. 0·023; P < 0·001) for treatments A, B and C, respectively. Within-pen variation in pig weight was similar for all treatments at day 14, 27 and 56 (P > 0·05), respectively. In experiment 2, daily gain was 403, 436 and 394 g/day (s.e. 13·0; P = 0·07) and FCE was 1·19, 1·16 and 1·24 g/g (s.e. 0·02; P < 0·05) during the period from day 0 to 26 for treatments A, B and C, respectively. The proportion of the diet selected as starter diet was 0·20, 0·57 and 0·53 (s.e. 0·024; P < 0·001) for treatments A, B and C, respectively. In experiment 3, daily gain was 465, 486 and 488 g/day (s.e.9·4; P > 0·05) and FCE was 1·14, 1·11 and 1·07 g/g (s.e. 0·015; P < 0·01) during the period from day 0 to 26 for treatments A, B and C, respectively. The proportion of the diet selected as starter diet was 0·21, 0·49 and 0·55 (s.e. 0·022; P < 0·001) for treatments A, B and C, respectively. Pig weight at day 49 was 36·4, 37·1 and 37·3 kg (s.e. 0·27; P = 0·09). It was concluded from these experiments that choice feeding did improve pig performance when a choice of starter and link diet was offered.

Potential environmental benefits of prospective genetic changes in broiler traits

A system approach-based Life Cycle Assessment (LCA) framework, combined with a simple mechanistic model of bird energy balance was used to predict the potential effects of 15 y prospective broiler breeding on the environmental impacts of the standard UK broiler production system. The year 2014 Ross 308 genotype was used as a baseline, and a future scenario was specified from rates of genetic improvement predicted by the industry. The scenario included changes in the traits of growth rate (reducing the time to reach a target weight 2.05 kg from 34 d to 27 d), body lipid content, carcass yield, mortality and the number of chicks produced by a breeder hen. Diet composition was adjusted in order to accommodate the future nutrient requirements of the birds following the genetic change. The results showed that predicted changes in biological performance due to selective breeding could lead to reduced environmental impacts of the broiler production chain, most notably in the Eutrophication Potential (by 12%), Acidification Potential (by 10%) and Abiotic Resource Use (by 9%) and Global Warming Potential (by 9%). These reductions were mainly caused by the reduced maintenance energy requirement and thus lower feed intake, resulting from the shorter production cycle, together with the increased carcass yield. However, some environmental benefits were limited by the required changes in feed composition (e.g., increased inclusion of soy meal and vegetable oil) as a result of the changes in bird nutrient requirements. This study is the first one aiming to link the mechanistic animal modeling approach to predicted genetic changes in order to produce quantitative estimates of the future environmental impacts of broiler production. Although a more detailed understanding on the mechanisms of the potential changes in bird performance and their consequences on feeding and husbandry would be still be needed, the modeling framework produced in this study provides a starting point for predictions of the effects of prospective genetic progress.

Constraints on energy intake in fish: the link between diet composition, energy metabolism, and energy intake in rainbow trout

The hypothesis was tested that fish fed to satiation with iso-energetic diets differing in macronutrient composition will have different digestible energy intakes (DEI) but similar total heat production. Four iso-energetic diets (2 × 2 factorial design) were formulated having a contrast in i) the ratio of protein to energy (P/E): high (H(P/E)) vs. low (L(P/E)) and ii) the type of non-protein energy (NPE) source: fat vs. carbohydrate which were iso-energetically exchanged. Triplicate groups (35 fish/tank) of rainbow trout were hand-fed each diet twice daily to satiation for 6 weeks under non-limiting water oxygen conditions. Feed intake (FI), DEI (kJ kg(-0.8) d(-1)) and growth (g kg(-0.8) d(-1)) of trout were affected by the interaction between P/E ratio and NPE source of the diet (P<0.05). Regardless of dietary P/E ratio, the inclusion of carbohydrate compared to fat as main NPE source reduced DEI and growth of trout by ~20%. The diet-induced differences in FI and DEI show that trout did not compensate for the dietary differences in digestible energy or digestible protein contents. Further, changes in body fat store and plasma glucose did not seem to exert a homeostatic feedback control on DEI. Independent of the diet composition, heat production of trout did not differ (P>0.05). Our data suggest that the control of DEI in trout might be a function of heat production, which in turn might reflect a physiological limit related with oxidative metabolism.

Measuring diet selection in dairy cows: effect of training on choice of dietary protein level

In a 7-week experiment, the ability of lactating Holstein-Friesian cows to select a consistent diet from two similar foods differing in calculated metabolizable protein to energy (MP/ME) yield was investigated. The effect on diet selection of training through previous access to foods separately, was measured. Food intake was recorded with 28 computer-linked feeders. All foods were mixtures of grass silage and concentrates. In week 1, all feeders contained a standard food. In weeks 2 to 7 a low protein food (LP) and a high protein food (HP) were offered in 14 feeders each. Group CHOICE had access to both foods as a choice from week 2. Group TR1 was trained by access to one food during days 8 to 10 and to the other during days 11 to 13. Group TR2 received the same training as group TR1 during days 8 to 13 which was repeated once during days 14 to 19. After training, TR groups had access to both foods as a choice. Groups LOPRO and HIPRO had only access to LP or HP, respectively in weeks 2 to 5 and to both foods as a choice in weeks 6 and 7. In weeks 2 to 5 LOPRO COWS consumed less dry matter and produced less milk than CHOICE or HIPRO cows. After a week of adaptation, untrained CHOICE COWS selected 662 (s.e. 27) g HP per kg of intake, a choice that differed significantly (P < 0·01) from random. In weeks 4 to 7 TR cows established similar diet choice: 696 (s.e. 21) g HP per kg intake and the proportion selected was unaffected by length of training. The between-day variation in diet choice within cows was not affected by treatment. It is concluded that, under the circumstances tested, training was not required for cows to distinguish between two mixed foods with different calculated MP/ME ratios and to select proportions significantly different from random.

Food choice and intake: towards a unifying framework of learning and feeding motivation

The food choice and intake of animals (including humans) has typically been studied using frameworks of learning and feeding motivation. When used in isolation such frameworks could be criticized because learning paradigms give little consideration to how new food items are included or excluded from an individual's diet, and motivational paradigms do not explain how individuals decide which food to eat when given a choice. Consequently we are posed with the question of whether individuals actively interact with the food items present in their environment to learn about their nutritional properties? The thesis of this review is that individuals are motivated to actively sample food items in order to assess whether they are nutritionally beneficial or harmful. We offer a unifying framework, centred upon the concept of exploratory motivation, which is a synthesis of learning and paradigms of feeding motivation. In this framework information gathering occurs on two levels through exploratory behaviour: (i) the discrimination of food from nonfood items, and (ii) the continued monitoring and storage of information concerning the nutritional properties of these food items. We expect that this framework will advance our understanding of the behavioural control of nutrient intake by explaining how new food items are identified in the environment, and how individuals are able to monitor changes in the nutritional content of their food resource.

A personal view of how ruminant animals control their intake and choice of food: minimal total discomfort

Voluntary food intake and the selection between foods are important subjects especially in ruminants in view of the economic importance of this class of animal and the complex digestive system with its attendant metabolic peculiarities. There is evidence that intake is limited by the capacity of the rumen as well as by metabolic factors; some theories assume that intake is controlled by the first limiting factor but this is not satisfying on physiological grounds and there is evidence that signals from feedback factors are integrated in an additive manner. It is now well established from research in which animals are given the chance to learn the metabolic consequences of eating food with a particular sensory profile, including a choice of foods, that animals including ruminants can adjust their diet, both quantitatively and qualitatively, to their nutrient requirements. It is proposed that they do this in order to minimise the total of the discomfort generated by the several signals from various body systems. The learning process is aided by the considerable day-to-day variation often seen in the intake of individual animals. An optimisation model is proposed and presented in a simple form, involving the addition of discomforts (calculated as the square of the deviation of the supply of metabolisable energy, crude protein and neutral-detergent fibre) and iterative elucidation of the intake at which total discomfort is minimal. With parameters appropriate for growing lambs the model provides reasonable agreement with observations, both in terms of daily intake and selection between foods of different protein contents. Manipulation of food composition and of nutrient requirements produces predictions broadly in agreement with reality except that protein deficiency has less severe consequences for the model than for real animals; it is proposed that protein deficiency be given more weighting than protein excess, and this may be true for other resources as well. This model is proposed as a philosophy and a starting point for further development and is not purveyed as a complete, working model. It nevertheless provides support for the concept of total minimal discomfort as a suitable base from which to view the control of intake and selection in all animals.

Predicting the effects of body fatness on food intake and performance of sheep

Adipose tissue produces signals that can have a profound effect on many physiological functions, including energy expenditure and food intake. The hypothesis that variation in food intake of sheep resulting from differences in animal fatness can be predicted from effects of animal fatness on energetic efficiency was subjected to three tests. First, an existing food intake model was adapted to account for effects of animal fatness, as estimated by condition score, on food intake. Parameter values were derived from data obtained with two of five treatment groups of an experiment where ewe lambs were fed either chopped hay or pelleted concentrates. The model predicted the intake of the remaining three treatment groups satisfactorily. The energy intake model was subsequently extended with a protein module based upon a Gompertz curve to simulate changes in body weight and condition score. The model predicted these changes satisfactorily for most treatment groups during the experimental period of 50 weeks. In a last test, the final body weights and body lipid contents of animals fed either hay or concentrates for a period of 3 years were predicted. The predictions for final body weight (77 or 118 kg) and lipid content in the empty body (26 or 58 %) were within the range of expectations for sheep with access to hay or concentrates, respectively. The biological implications of the hypothesis that body fatness acts upon voluntary intake via its effects on energetic efficiency are discussed.

Control of energy balance in relation to energy intake and energy expenditure in animals and man: an ecological perspective

In this paper, we consider the control of energy balance in animals and man. We argue that patterns of mammalian feeding have evolved to control energy balance in uncertain environments. It is, therefore, expected that, under sedentary conditions in which the diet is rich in nutrients and abundantly available, animals and man will overeat. This suggests that no physiological defects are needed to induce overweight and ultimately obesity in man. Several considerations arise from these observations. The time period over which energy balance is controlled is far longer than allowed by most experiments. Physiological models of energy balance control often treat excess energy intake as a defect of regulation; ecological models view the same behaviour as part of normal energy balance control in environments where resources are uncertain. We apply these considerations to common patterns of human and animal feeding. We believe that the ecological perspective gives a more accurate explanation for the functionality of excess fat and the need to defend nutrient balance and avoid gross imbalances, as well as explaining hyperphagia in the face of plenty. By emphasising the common features of energy balance control in different mammalian species, the importance of changes in behaviour to accommodate changes in the environment becomes apparent. This also opens up possibilities for the control of body weight and the treatment of obesity in man.

Body fatness affects feed intake of sheep at a given body weight

In a 1-yr experiment, nutritional treatments were used to produce different combinations of BW and BCS in lambs. The experiment served to quantify the effects of BW and BCS on ADFI by sheep. Ewe lambs (n = 78) were assigned to treatment groups that had ad libitum access to one feed at a time. Three feeds were used: a medium-quality chopped hay (L), a pelleted feed based on oat feed (M), and a pelleted feed based on barley (H). Three groups received only one of these feeds throughout. Two groups first received H and then were switched to M when they reached a BW of 45 or 65 kg. Two groups first received L and then were switched to M or H after reaching a BW of 45 kg. Three groups first received H or M but were switched to L after reaching a BW of 45, 65, or 95 kg. Daily feed intake, BW, and BCS were recorded, and ME content of the feeds was estimated in a separate digestibility experiment. The lambs consuming M ate more (P < 0.001) feed than lambs consuming H, but this had no significant effects on ME intake or gain in BW or BCS. Animals that had had access to L were lean for their BW when switched to H or M and showed compensatory intake and gain. Animals switched from M or H to L all lost BCS; BW change depended on the BW at the switch. The treatments produced different combinations of BW and BCS for animals with access to the same feed. The ADFI of a given feed varied systematically with BCS for animals of a given BW. The model ADFI = a x BW x [1 - (b x BCS)] gave a reasonable description of the data in all treatments. A model using BW, BCS, and their interaction gave a slightly better fit but explained little more of the variation in ADFI than the simpler model. The implications of the collected data are that BW alone is an insufficient descriptor of the animal to correctly predict feed intake and that intake predictions can be improved by taking BCS into account.

A dynamic model of metabolizable energy utilization in growing and mature cattle. III. Model evaluation

Component models of heat production identified in a proposed system of partitioning ME intake and a dynamic systems model that predicts gain in empty BW in cattle resulting from a known intake of ME were evaluated. Evaluations were done in four main areas: 1) net efficiency of ME utilization for gain, 2) relationship between recovered energy and ME intake, 3) predicting gain in empty BW from recovered energy, and 4) predicting gain in empty BW from ME intake. An analysis of published data showed that the net partial efficiencies of ME utilization for protein and fat gain were approximately 0.2 and 0.75, respectively, and that the net efficiency of ME utilization for gain could be estimated using these net partial efficiencies and the fraction of recovered energy that is contained in protein. Analyses of published sheep and cattle experimental data showed a significant linear relationship between recovered energy and ME intake, with no evidence for a nonlinear relationship. Growth and body composition of Hereford x Angus steers simulated from weaning to slaughter showed that over the finishing period, 20.8% of ME intake was recovered in gain. These results were similar to observed data and comparable to feedlot data of 26.5% for a shorter finishing period with a higher-quality diet. The component model to predict gain in empty BW from recovered energy was evaluated with growth and body composition data of five steer genotypes on two levels of nutrition. Linear regression of observed on predicted values for empty BW resulted in an intercept and slope that were not different (P < 0.05) from 0 and 1, respectively. Evaluations of the dynamic systems model to predict gain in empty BW using ME intake as the input showed close agreement between predicted and observed final empty BW for steers that were finished on high-energy diets, and the model accurately predicted growth patterns for Angus, Charolais, and Simmental reproducing females from 10 mo to 7 yr of age.

A lifetime perspective on foraging and mortality

Food intake carries many potential risks which may impair an animal's reproductive success not only in the current breeding cycle, but also for the rest of its lifetime. We examine the lifetime trade-off between the costs and benefits of food intake by presenting a simple animal foraging model, where each unit of food eaten carries with it a risk of mortality. We show that the optimal food intake rate over an animal's lifetime, for both semelparous and iteroparous animals, is not maximal. Instead, animals are required to strike a balance between the immediate reproductive benefits of gathering food and the future reproductive costs incurred by the food's mortality risk. This balance depends upon the lifespan of the animal as well as the nature of the risk. Different mortality risks are compared and it is shown that a mortality risk per unit time spent foraging is not, in general, equivalent to a mortality risk per unit of food consumed. The results suggest that a mortality risk per unit of food consumed, such as that presented by the presence of a toxin or of a parasite in the diet, has important consequences for feeding behaviour and is a possible factor involved in food intake regulation.

Theoretical developments in the study and prediction of food intake

The purpose of the present paper is to review recent theoretical developments in food intake modelling applied to animal science and ecology. The models are divided into those that have been developed for intensive agricultural systems, and those which consider more extensive systems and natural systems. For the most part the present paper discusses models that predict the food intake of herbivores. The mechanisms of each model are discussed, along with a brief mention of the experimental support for the most popular models. We include a discussion of models that approach the study of food intake behaviour from an evolutionary perspective, and suggest that lifetime models are especially useful when food intake carries an intrinsic cost. These long timescale evolutionary models contrast with the more common food intake models, whose timescale is usually much shorter. We conclude that the 'eating to requirements' model highlights an important food intake mechanism that provides an accurate predictive tool for intensive agricultural systems. The mechanisms of food intake regulation in extensive systems are less certain, and closer links between the ideas of animal science and ecology will be helpful for improving our understanding of food intake regulation.

Consequences of genetic change in farm animals on food intake and feeding behaviour

Selection in commercial populations on aspects of output, such as for growth rate in poultry. against fatness and for growth rate in pigs, and for milk yield in cows, has had very barge effects on such outputs over the past 50 years. Partly because of the cost of recording intake, there has been little or no selection for food intake or feeding behaviour. In order to predict the effects of such past, and future, selection on intake it is necessary to have some suitable theoretical framework. Intake needs to be predicted in order to make rational feeding and environmental decisions. The idea that an animal will eat 'to meet its requirements' has proved useful and continues to be fruitful. An important part of the idea is that the animal (genotype) can be described in a way that is sufficient for the accurate prediction of its outputs over time. Such descriptions can be combined with a set of nutritional constants to calculate requirements. There appears to have been no change in the nutritional constants under selection for output. Under such selection it is simplest to assume that changes in intake follow from the changes in output rates, so that intake changes become entirely predictable. It is suggested that other ways that have been proposed for predicting intake cannot be successful in predicting the effects of selection. Feeding behaviour is seen as being the means that the animal uses to attain its intake rather than being the means by which that intake can be predicted. Thus, the organisation of feeding behaviour can be used to predict neither intake nor the effects of selection on it.

Conditioned food aversions: principles and practices, with special reference to social facilitation

Conditioned food aversion is a powerful experimental tool to modify animal diets. We have also investigated it as a potential management tool to prevent livestock from grazing poisonous plants such as tall larkspur (Delphinium barbeyi), white locoweed (Oxytropis sericea) and ponderosa pine (Pinus ponderosa) on western US rangelands. The following principles pertain to increasing the strength and longevity of aversions: mature animals retain aversions better than young animals; novelty of the plant is important, although aversions can be created to familiar plants; LiCl is the most effective emetic, and the optimum dose for cattle is 200 mg/kg body weight; averted animals should be grazed separately from non-averted animals to avoid the influence of social facilitation which can rapidly extinguish aversions. Social facilitation is the most important factor preventing widespread application of aversive conditioning. When averted animals see other animals eat the target food they will sample it, and if there is no adverse reaction they will continue eating and extinguish the aversion. However, if averted animals can be grazed separately, aversions will persist. Aversive conditioning may provide an effective management tool to prevent animals from eating palatable poisonous plants that cause major economic loss.

Does the study of feeding behaviour benefit from a teleonomic framework?

In this paper we respond to the criticisms of Provenza et al. (1998) that our framework of learning and feeding motivation (Day et al. 1998) resorts to higher order goals, which cannot be falsified by experimentation. We assert that in order to be able to predict the feeding behaviour of animals we first need to understand what they are trying to achieve (i.e. invoke teleonomy). We then detail our framework in such terms that one could envisage experiments that could quantitatively test its predictions. We contend that the framework of 'the self-organization of behaviour' proposed by Provenza et al. (1998) cannot lead to such quantitative predictions, since it is invoked to describe feeding behaviour of animals a posteriori. It is our own desire, by contrast, to assess feeding behaviour a priori, which leads us to propose and defend our framework of learning and feeding motivation.

Food intake and diet selection in sheep: the effect of manipulating the rates of digestion of carbohydrates and protein of the foods offered as a choice

An experiment was designed to investigate whether the degree of synchrony between the rates of digestion of carbohydrates and N of foods offered as a choice would have an effect, through their consequences, on the short- and long-term diet selection of sheep. Four foods (RL, RH, SL and SH) with the same high metabolizable energy, and similar high metabolizable protein contents were made into pellets. Foods RL and RH were based on a rapidly fermentable carbohydrate source and foods SL and SH on a slowly fermentable carbohydrate source; within each source one food (RL or SL) had a low, and the other (RH or SH) a high, rumen-degradable protein (RDP) content. The foods within a carbohydrate source were offered either singly or as a choice (RL/RH or SL/SH) to eleven rumen-fistulated mature sheep. The design was two 3 x 3 Latin squares (replicated once) with 5-week periods; squares consisted of two single foods and their respective choice. Weeks 1, 3 and 5 were considered to be controls, and weeks 2 and 4 used for rumen infusions of either urea or fructose infused over 4 h (10.00-14.00 hours). Food intake (FI) and diet selections (DS) were recorded daily and every 2 h (08.00-16.00 hours) on days 2-5 of each week; rumen pH and NH3 concentrations were also measured during these time intervals of day 5. As expected, feeding treatment affected significantly the rumen measurements: rumen NH3 concentrations were higher on foods RH and SH, and rumen pH lowest on RL. Daily FI was lowest on treatments SL, and choice SL/SH. The mean daily proportion of the low-RDP food in the selected diet was lower when the carbohydrate source was rapidly (choice RL/RH) rather than slowly fermentable (choice SL/SH); this was consistent with the experimental hypothesis. Short-term infusions affected further rumen variables (in the expected directions), irrespective of feeding treatment. However, DS over the 4h infusion period were unaffected; these short-term DS were consistent with the ones selected over the longer term (daily). The results suggest that the long-term (daily) diet selection of sheep may be affected by the degree of synchrony of energy and protein to the rumen. The fact that diet selections were not altered further by short-term manipulations of these supplies might reflect inadequacies of the methodology (infusions) adopted here.