The main aim of the work described in the present thesis was to examine the consistency of individual differences in behavioural and physiological responses to acute stressors in dairy cattle in a longitudinal fashion, i.e. both during rearing and in adulthood.
Information about the variation in the reactivity of dairy cattle to a “real-life” challenge was provided by a study of heifers’ responses to first-time machine milking. A pharmacological validation experiment helped to interpret the response patterns that calves exhibited in behavioural tests designed to induce stress. The long-term longitudinal study enabled clarification of the relationship between responses of heifers to first machine milking and their earlier responses to behavioural tests.
In this general discussion, I will consider the results in a wider context, and also specifically focus on (i) the multidimensional nature of response patterns to stress, (ii) the significance of individual differences in stress responsiveness in (farm) animals, and (iii) some practical implications of the findings of this thesis.
MULTIDIMENSIONAL RESPONSE PATTERN TO STRESS
The dominant picture provided by the results of this study is that the responsiveness of dairy cattle (both cows and calves) to challenge is mediated by multiple traits, i.e., is multidimensional. Applying principal component analysis (PCA) to behavioural and physiological response measures always yielded at least two components (dimensions) without cross-loading of variables, even when it was widely assumed that those measures could be related to the same trait. This seems to contrast with prevailing concepts of individual differences which emphasized the existence of major unifying dimensions such as, for example, fearfulness (Boissy, 1995, Jones, 1996), temperament (Clarke and Boinski, 1995), behavioural syndrome (Sih et al., 2004a, b) or coping style (Koolhaas et al., 1999, see Korte et al., 2005 for a related characteristic defining the unidimensional distinction between “hawk” and “dove” type personalities in animals and men). The basic premise of these concepts is that animals exhibit consistent individual differences in behavioural and physiological responses across a wide range of different environmental challenges. However, from many reports in the literature, covering many species, it appears that multidimensional response patterns to stressors seem to be the rule rather than the exception. For example, multiple independent characteristics were obtained after correlational analyses, including PCA, of behavioural and physiological measures of response to challenge in sunfish (Coleman and Wilson, 1998), geese (Kralj-Fiser et al., 2006), quail (Mignon-Grasteau et al., 2003; Miller et al., 2006), deer (Pollard et al., 1994; Bergvall et al., 2011), laboratory rodents (Kanari et al., 2005; Ibanez et al., 2007, 2009), dogs (Svartberg, 2005), pigs (Forkman et al., 1995; Spoolder et al., 1996; Mendl et al., 1998; D’Eath and Burn, 2002; Janczak et al., 2003a, b; Van Erp – van der Kooij et al., 2002; Brown et al., 2009), and cattle (Piovezan et al., 1998; Kilgour et al., 2006; Petherick et al., 2009a). Clearly, there is a need for alternative interpretations of individual differences that go beyond univariate classifications.
Domain-specificity of response
Using the temperamental trait “boldness” (or the “shy-bold continuum”) as an example, Wilson and colleagues (Wilson et al., 1994; Wilson, 1998) argued that, depending on the ecological circumstances, it might be favourable for an individual to facultatively express either a shy or a bold reaction pattern in the face of challenge (e.g., novelty, a predator, a conspecific, etc.). This was described as “domain specificity” or “phenotypic plasticity” of response, as opposed to “domain generality” or “phenotypic rigidity” (see also Réale et al., 2000; Sih et al., 2004a, b). A similar context dependency of response was proposed for characteristics like fearfulness, anxiety, or emotionality (Archer, 1979; Ramos and Mormède, 1998). This might explain, for example, why the same animal is fearful of humans, but non-fearful of novel objects (e.g., Visser et al., 2003; Janczak et al., 2003a; Gibbons et al., 2009a), or aggressive and agitated towards conspecifics, but non-aggressive and non-agitated during manual restraint or handling (e.g., Mendl et al., 1998; Réale et al., 2000; D’Eath and Burn, 2002), etc. The current findings in dairy cattle, however, show that different dimensions of responsiveness not so much differed with regard to the context (i.e., the test) in which behavioural and/or physiological measures were obtained, but that each dimension largely reflected the same measure recorded in different tests, i.e., locomotion, vocalisation or correlated behavioural and adrenocortical responses to novelty (see chapters 2, 3 and 6). Explaining the present results only in terms of context specificity of response would, therefore, require a rather complex line of argument because it would not only mean that the same underlying trait (for example, fearfulness) is differentially expressed (i.e., by vocalisation, by locomotion, or by correlated behavioural and adrenocortical responses to novelty) depending on the context (i.e., different tests, or different aspects of the same test such as novelty or social isolation), but also that different forms of context dependent fear responding are simultaneously exhibited in the same animal (see chapter 2). Thus, it is worthwhile to consider other frameworks too.
Qualitative and quantitative dimensions of responsiveness to challenge
Another explanation more closely considers the nature of two main concepts of individual differences: coping style and underlying fearfulness. The concept of coping style assumes that individuals may show alternative types of response patterns to the same challenge, e.g., either a passive or an active type of reaction (Koolhaas et al., 1997, 1999). Importantly, passive coping by no means indicates that animals are “passive” in the sense that they do not respond. On the contrary, a passive coping strategy is thought to include, for example, enhanced parasympathetic activity and, under some circumstances, a high reactivity of the hypothalamo-pituitary-adrenocortical (HPA) axis (Koolhaas et al., 1999; Korte et al., 2005), and it was therefore suggested to rename the word “passive” into “reactive” (Koolhaas et al., 1997). The concept of fearfulness, in contrast, makes a distinction between, on the one hand, fearful animals that are highly emotionally aroused by a challenging situation and, hence, exhibit activation of neuroendocrine systems involved in stress responsiveness (such as the HPA axis and the sympatho-adrenomedullary system) and, on the other, non-fearful or “relaxed” animals who do not perceive that same situation as stressful or alarming and, therefore, do not show any enhanced, or at least show less elevated, biological responses (Boissy, 1995; Jones, 1996). Thus, coping style seems to reflect the type of response an animal makes (i.e., how an animal reacts), and fearfulness indicates the level of responsiveness to challenge (i.e., how strongly an individual reacts). From a conceptual point of view, coping style would then represent the qualitative dimension of the stress response, and fearfulness the quantitative one. In the present thesis, this twodimensional or “two-tier” model (see Koolhaas et al., 2007; Coppens et al., 2010) was proposed to explain the multivariate response pattern of calves to open field (OF) and novel object (NO) tests (chapter 3). Correlated behavioural and adrenocortical responses of calves to novelty were suggested to reflect underlying fearfulness, while OF locomotion was thought to be mediated by a type of coping style.
The assumptions of this model were (i) that in the NO test paradigm all fearful calves adhered to the same behavioural strategy of response (i.e., they had long latencies to first contact with the NO, and spent short times in contact with it), (ii) that the level of responsiveness (i.e., the level of avoidance of the NO, and the strength and amplitude of plasma cortisol responses), was higher in fearful calves than in non-fearful ones, and (iii) that during the OF test, both fearful and non-fearful calves were equally likely to consistently exhibit either high or low levels of locomotion, possibly under the influence of coping style. These assumptions would predict that the administration of a fear-reducing pharmacological agent to calves prior to OF and NO tests would decrease the strength of the fear response (i.e, decrease avoidance of the NO and, hence, increase their interaction with it and decrease the magnitude of the cortisol response), but would not affect OF locomotion. These predictions were largely confirmed in the pharmacological validation study (chapter 4), thereby lending important experimental support to the model. A crucial implication of this model would be that the nature of the experimental paradigm (e.g., the type of behavioural test) that is used to assess individual differences in responsiveness to challenge determines the extent to which fearful animals will exhibit similar or, under the influence of coping style, differential response patterns to the same challenge. If we accept the argument that some paradigms, or some behavioural or physiological measures recorded within the same paradigm, primarily reflect the level of responsiveness (e.g., fearfulness), and others mainly indicate the qualitative type of response (e.g., coping style), then by definition the resulting overall response pattern should be multidimensional if both types of paradigms are used, or both types of measures are recorded, in the same experiment (see Ramos and Mormède, 1998). A more conceptual version of the two-dimensional model of responsiveness previously presented in chapter 3 is shown in Figure 7.1. The horizontal axis represents the quantitative dimension of the response pattern to challenge, which encompasses the strength and the amplitude of the response. These may vary between high (left-hand side) and low (right-hand side). The horizontal axis represents the qualitative dimension of the response, with animals on both extremes of the distribution differing in, for example, the type of coping style. (e.g., reactive versus proactive).
I suggest that a closer inspection of previous papers, in particular those on the species most frequently used to examine coping style (i.e., laboratory rodents and pigs), provides further support for multidimensionality due to the dissociation between “level” and “type” of response. For example, Koolhaas et al. (2001) reported that the attack latency of wild type male rats in a resident intruder test (an index of coping style in rats and mice, see Benus et al., 1991) was unrelated to the time spent in the open arm of the elevated plus maze (an index of fear and anxiety, see Pellow et al., 1985). Correspondingly, Sgoifo et al., (1996) found that the attack latency of rats was significantly associated with plasma catecholamine concentrations after social (defeat experience) or non-social (exposure to a shock prod in the home pen) stress (reflecting the level of activation of the sympathetic nervous system, a neuroendocrine index of coping style, see Koolhaas et al., 1999; Korte et al., 2005), but unrelated to stressinduced plasma corticosterone (which has been suggested to reflect the level of fear or anxiety, e.g., Landgraf and Wigger, 2003; Marquez et al., 2005). Similarly, in pigs, putative measures of coping style, such as the behavioural response to a backtest or the level of aggression during social confrontations with conspecifics (see Hessing et al., 1993; Ruis et al., 2000; Bolhuis et al., 2004, 2005), were mostly unrelated to putative measures of fearfulness or anxiety, such as the latency to approach a novel object or a human (e.g., Jensen, 1994, Jensen et al., 1995a; Forkman et al., 1995; Spoolder et al., 1996; Van Erp – van der Kooij et al., 2002; Janczak et al., 2003b; Brown et al., 2009).
On the assumption that a two-dimensional model of responsiveness based on the interaction between the level of responsiveness and the qualitative type of response is valid, one could surmise that an emotional state of fear – which would be most apparent in animals situated on the right-hand side of the two-dimensional distribution presented in Figure 7.1 – is a prerequisite for the expression of individual differences in coping style. Studies of mice (Sluyter et al., 1996) and piglets (Erhard et al., 1999) favoured this idea by suggesting that animals only differed in putative behavioural measures of coping style (i.e., defensive burying or tonic immobility in mice and piglets, respectively) when they experienced the test situation as stressful or aversive (i.e., when mice were tested with fresh bedding, or when piglets exhibited behavioural signs of emotional distress in response to handling). Steimer et al. (1997) and Koolhaas et al. (2007), on the other hand, proposed two-tier models of responsiveness, similar to the model in Figure 7.1, where differences in coping style might also be expressed under low fear circumstances, for example in differential levels of impulsiveness. In the current thesis it was suggested that in non-fearful animals OF locomotion might be controlled by characteristics other than coping style, such as the tendency to explore (chapters 3), but this needs further research. It would, however, also be perfectly conceivable that a coping style trait in dairy cattle mediates both the need to explore in non-fearful animals as well as the expression of locomotion (i.e., either fear-induced immobility or fear-induced activity) in fearful ones.
Figure 7. 1 Model of responsiveness of animals to challenge, based on two independent (orthogonal) dimensions: a quantitative dimension (horizontal axis) representing features like the strength and amplitude of the response, and a qualitative dimension (vertical axis) reflecting the qualitative type of reaction. Animals can be distributed in the two-dimensional space according to their behavioural and physiological response pattern to various challenges. Animals on the far left-hand side of the quantitative dimension experience low emotional arousal, and have baseline states of the neuroendocrine systems involved in the stress response, whereas animals on the right-hand side of the dimension experience high levels of fear or emotional distress, and exhibit increased activity of neuroendocrine response systems in reaction to the same challenge. Animals situated on the extremes of the qualitative dimension exhibit differences in, for example, type of coping style (e.g., reactive versus proactive) or nervous system balance (e.g., high parasympathetic versus high sympathetic activity). Animals are assumed to exhibit individual differences in the level of responsiveness, and/or in the qualitative type of response, depending on the nature of the (experimental) challenge or paradigm they are subjected to.
It is important to note that neither context specificity of response, nor the recognition of qualitative and quantitative dimensions of fear responses to stressors, seemed sufficient to fully explain the present response patterns observed in dairy cattle. Underlying sociality (or sociability), i.e. the motivation to remain close to conspecifics (Erhard and Schouten, 2001; Sibbald et al., 2006; Gibbons et al., 2010) was proposed as a separate trait underpinning the responsiveness of calves and cows to the challenges examined here (chapters 2, 3 and 6). This trait is not only related to social separation distress but also to social bonding and the capacity to be comforted by peers, which are regulated by distinct reward areas in the brain (e.g., Massen et al., 2010). This explains why putative behavioural measures of underlying sociality in cattle, (e.g. vocalisations) recorded during social isolation or separation, were positively correlated with the duration of non-aggressive social interactions (including sniffing and licking) with conspecifics (Boissy and Bouissou, 1995; Boissy and Le Neindre, 1997). In the present thesis, it was hypothesized that the rate of OF vocalisation reflects sociality in dairy cattle (chapters 2, 3 and 6). This behavioural measure was largely uncorrelated with other behavioural and physiological measures. Therefore, the two-dimensional model presented in Figure 7.1 should be extended with at least one additional dimension, representing underlying sociality, perpendicular to the other two axes (see also chapter 3). For the time being, I assume that underlying sociality behaved like a quantitative trait, i.e., that the number of vocalisations reflected the strength of the response, with high and low levels of vocalisation referring to high and low levels of sociality, respectively. Theoretically, however, qualitative dimensions of sociality cannot be completely ruled out, but remain speculative.
The concept of multidimensionality of response patterns to challenge is also supported by (molecular) genetic studies. Gutierrez-Gil et al. (2008), for example, searched for genomic regions (i.e., quantitative trait loci or QTLs) influencing behavioural traits pertaining to cattles’ response patterns exhibited during exposure to social separation and flight distance tests. Notably, QTLs associated with traits assessed in different tests did not overlap, suggesting that different aspects of cattle “temperament” or “fearfulness” are controlled by different underlying genetic factors. Earlier work in quail (Mills and Faure, 1991; Faure and Mills, 1998) has revealed that genetic lines can be created showing either high or low levels of social reinstatement behaviour (a measure of sociality), or long or short durations of tonic immobility (a commonly used measure of fearfulness). This suggests that sociality and fearfulness are indeed independent traits, and supports the current hypopthesis that these traits are also independently expressed in dairy cattle (see chapters 2, 3 and 6). Conversely, extensive studies in genetic lines of a wild bird species, the great tit (Parus major), indicated that behavioural profiles may consist of a range of genetically correlated traits, including risk taking behaviour, aggression, exploration, and interaction with a novel object, which would argue in favour of the existence of common sets of genes that exert an influence on multiple characteristics that shape the way these birds cope with environmental challenge (Van Oers et al., 2004; Carere et al., 2005; Groothuis and Carere, 2005). Under natural circumstances, however, selection pressure may also act on multiple independent traits simultaneously, producing behavioural syndromes comprised of phenotypically rather than genetically correlated traits (e.g., Sih et al., 2004a, b; Réale et al., 2007; Wolf and Weissing, 2010). Moreover, behavioural and physiological differences between genetic lines should be treated with care even under controlled experimental conditions. Such differences may be the result of fortuitous or accidental co-selection of genetically unrelated traits, and, therefore, systematic breeding experiments are required to identify true genetic links (see Castanon and Mormède, 1994; Mormède et al., 1994; Ramos and Mormède, 1998; Groothuis and Carere, 2005). The use of high-low sampling from a phenotypic distribution, as a tool to create groups with divergent response profiles (e.g., Ruis et al., 2000; Geverink et al., 2004 ), may introduce similar risks, in particular when different independent traits are not equally distributed among the population of interest.
Multidimensionality of temperament across species
The present suggestion that responsiveness to challenge in dairy cattle is mediated by multiple independent traits, including fearfulness, sociality and coping style, is supported by a considerable body of literature. Multiple traits (e.g., the “Big Five”) are also assumed to constitute human personality (e.g., Zuckerman, 1991; Funder, 2001), and a multidimensional approach is increasingly advocated in the context of research into animal personality, behavioural syndromes or coping style (e.g., Erhard and Schouten, 2001; Gosling, 2001; Visser, 2002; Sih et al., 2004a, b; Koolhaas et al., 2007), even to the point where it was proposed that animal temperament actually consists of five categories of traits (some of which may or may not be genetically linked), including “aggressiveness”, “avoidance of novelty”, “willingness to take risks”, “exploration”, and “sociability” (Réale et al., 2007). I argue that the application of a multidimensional model of responsiveness, based on the fundamental distinction between qualitative and quantitative dimensions of behavioural and physiological responses of animals to challenge, could be one way of reconciling the concept of fearfulness with that of coping style. For one thing, this model may help to explain multidimensional response patterns without questioning the validity of either concept, as has happened previously (e.g., in the case of coping style in farm animals, particularly pigs, see Jensen, 1995; Jensen et al., 1995b; Spoolder et al., 1996; Blokhuis et al., 2001). At the same time, in contrast with suggestions made by some authors (e.g., Jones, 1996; Landgraf and Wigger, 2003; Carere et al., 2005), this model would question the idea that coping style and fearfulness are similar or the same traits.
SIGNIFICANCE OF INDIVIDUAL DIFFERENCES
Individual differences as a consequence of natural selection
It is becoming increasingly clear that animals in the wild are exposed to selection pressures that facilitate the development and maintenance of consistent individual differences in behaviour within the same population, and that trade-offs play a key role in this respect (see Sih et al., 2004a, b; Stamp, 2007; Wolf et al., 2007; Wolf and Weissing, 2010). For example, a trade-off may exist between current and future growth and reproduction, and, hence, some animals may adopt a behavioural strategy that maximizes growth and reproduction in the short term, whereas others may consistently behave in such a way that growth and reproduction are safeguarded in the long run. The former category of animals may exhibit “bold” and aggressive personality traits, whereas “shy” and more cautious personality traits may prevail in the latter category (e.g., “risk takers” versus “risk avoiders”, see Wilson, 1994). Within each “life history strategy”, the animal strikes a balance between benefits and costs, i.e., “bold” animals have a fast growth and a high reproductive rate at the cost of higher risks of injury and mortality due to predation, and “shy” animals exhibit slower growth and lower reproductive rates, but at the same time their risk of injury and mortality is also lower (see also Korte et al., 2005). Consequently, different personality types may ultimately obtain equal overall fitness in terms of gene preservation. Similarly, there may be trade-offs between adaptation to different social and environmental conditions in migratory species. For example, in wild house mice aggressive animals are the most successful under stable territorial conditions, whereas non-aggressive ones thrive during emigrations and the establishment of new territories (Van Oortmerssen and Bussen, 1989): these types are believed to represent animals with a proactive (active) or a reactive (passive) coping strategy, respectively (Benus et al., 1991; Koolhaas et al., 1997, 2001). Thus, under natural circumstances, when animals of the same population are faced with the same challenging situation, consistent individual behavioural differences may be linked to profound differences in health and other life outcomes such as reproductive success (see Dingemanse and Réale, 2005; Korte et al., 2005; Réale et al., 2007). Behavioural differences of this sort have therefore been referred to as “adaptive personality differences” (e.g., Dingemanse and Wolf, 2010; Wolf and Weissing, 2010).
Biological basis of individual differences and temperament
Comprehensive animal models are beginning to unravel the neurobiological, immunological and genetic factors and mechanisms underlying the relationship between personality and health, and noticeable relations between behavioural characteristics and aspects of immunocompetence and disesase susceptibility have been reported (e.g., Landgraf and Wigger, 2003; Cavigelli, 2005; Korte et al., 2005; Kavelaars and Heijnen, 2006; Koolhaas et al., 2006; Koolhaas, 2008; Salome et al., 2008). For instance, individual differences in behavioural measures of coping style in mice and rats, such as the attack latency in a resident intruder test, are associated with differences in the propensity to develop stress-related pathologies (e.g., hypertension and atherosclerosis), and in the susceptibility to autoimmune disease (Koolhaas, 1994; Kavelaars et al., 1999). Another consistent behavioural characteristic in rats, locomotor activity in an open field, correlated with the propagation of injected tumor cells and with the progression of experimentally induced arthritis (Sajti et al., 2004a, b). Divergent genetic selection in rats for a neurobiological characteristic, i.e., sensitivity to the dopaminergic agonist apomorphine, resulted in lines that differ in a wide range of biological variables, including behavioural and physiological responses to novel and social challenges, and immunological reaction patterns to inflammatory (autoimmune) and infectious diseases (Cools et al., 1990, 1993; Kavelaars et al., 1997, Teunis et al., 2004). These studies provided confirmation of the now widely accepted notion that intricate and reciprocal relations exist between the central nervous, endocrine, and immune systems (see Ader et al., 1995; Glaser and Kiecolt-Glaser, 2006). Personality traits may then be viewed as fundamental moderating or intervening variables that, in a hierarchical sense, operate at a high level in the body (i.e., in the brain), and affect both the intensities and the types of a wide range of biological responses the individual mounts during challenge (see Boissy, 1995; McEwen, 2001; Korte et al., 2005).
Individual differences in domestic animals
Comparative research shows that similar (co)variations in behavioural and physiological response patterns to challenge exist in a wide range of vertebrate species, from fish to mammals, including humans (e.g., Wilson, 1994; Gosling, 2001; Sih et al., 2004a, b; Øverli et al., 2007; Mehta and Gosling, 2008). This suggests that similar “adaptive personality” traits have been conserved during evolution across species, involving common biological (e.g., neural and neuroendocrine) substrates, and, possibly, homologous genes (Flint et al., 1995; Mormède et al., 2002; Øverli et al., 2007). In comparison with their wild ancestors, domestication has undoubtedly altered the physical appearance and the behaviour of farm animals in many ways. There is, however, reasonable consensus that most of these changes are quantitative rather than qualitative in nature, or, in genetic terms, that during domestication no genes have disappeared from the gene pool (Price, 1999; Jensen, 2001; Mignon-Grasteau et al., 2005). With regard to behavioural and physiological responses to challenge, this means that it is the amplitude and vigour of response that has mainly changed (i.e., mostly decreased in the course of domestication) rather than the variability and diversity in response repertoire. In other words, there is reason to assume that even in the absence of ancient selection pressures that originally shaped the emergence of “adaptive personality” differences in animals and man (see above), modern farm animals still harbour a considerable potential to express differential behavioural and physiological response patterns to challenge, and, correspondingly, experience different fitness (e.g., health) consequences under divergent environmental conditions. Numerous examples in the scientific literature substantiate the above assumption. Studies in pigs demonstrated that (early) differences in the behavioural response to the backtest (a putative measure of coping style) or to other behavioural tests intended to provoke individual temperamental differences, were associated with differences in, for example, responses of the immune system to experimental antigenic challenges (Hessing et al., 1995; Bolhuis et al., 2003), growth, lean meat percentage and meat quality (Hessing et al., 1994; Van Erp – van der Kooij et al., 2000, Brown et al., 2007), maternal behaviour and reproductive success (Spinka et al., 2000; Thodberg et al., 2002a, b; Janczak et al., 2003a), and the performance of abnormal stereotypic behaviours (Geverink et al., 2003). Likewise, variation in struggling during restraint in a chute (crush), or in flight speed (i.e., the time taken for an individual animal to cover a set distance when it is released from a chute or weighting crate), recorded in beef cattle prior to the fattening or service period, covaried with variation in later measures of growth and meat quality (Voisinet et al., 1997a, b; Petherick et al., 2002, 2009b; King et al., 2006), immune function (Fell et al., 1999; Oliphint et al., 2006), and pregnancy rate (Cooke et al., 2009). Behavioural temperament of Limousin heifers assessed during standard encounters with a human handler was genetically correlated with maternal behaviour (i.e., licking the new born calf) and fertility and calving rate (Phocas et al., 2006). In poultry, differences in a range of neuroendocrine and behavioural response patterns to brief experimental challenges were related to (subsequent) differences in growth (Marin et al., 1999, 2003), egg production (Uitdehaag et al., 2008a), and the propensity to engage in (potentially harmful) feather pecking behaviours directed to conspecifics (Van Hierden et al., 2002; Rodenburg et al., 2004; De Haas et al., 2010).
Individual differences in dairy cattle
So far, in comparison with other farm animal species, less research has been devoted in dairy cattle to the consequences of individual differences in stress responsiveness for the capacity of the animals to adapt to (challenging factors in) their actual living environment. Nevertheless, the available data seem to agree with the general picture emerging from studies in other species (see above). In Normandy cows (a French dairy breed), for example, behavioural and heart rate responses to unfamiliar test situations correlated with muscle characteristics, including temperature and pH, recorded at the slaughter house three weeks later (Bourguet et al., 2010). Schrader (2000) examined relationships in dairy cows between (i) behavioural reactions to brief experimental challenges, including “tail fixation”, (ii) measures of spontaneous activity in the home environment (cubicle house), calculated from 24-hour recordings of walking, standing and lying behaviour over a 4-day period, (iii) a measure of “regularity” of home pen behaviour, reflecting the day-to-day consistency of dairy cow activity in terms of the location in the barn (walking area, open yard, cubicles, feeding rack) where each animal was at each 5-min interval across the day, and (iv) an “agonistic index”, indicating for each individual cow the level of success during agonistic interactions. Significant and mutual correlations were found between the behavioural response (i.e., the number of steps and kicks) to tail fixation, the day-to day behavioural regularity, the average duration of lying bouts, and the agonistic index. It is tempting to speculate from this work that (consistency of) home pen activity and aggression in dairy cows are mediated by a common underlying predisposition. Indeed, there is recent evidence that aspects of both home pen activity and aggressive behaviour in dairy cows possess trait-like qualities (Müller and Schrader, 2005; Gibbons et al., 2009b). The potential relevance of such a trait is further highlighted by results of Galindo et al. (2000), showing that an index of “displacements” (reflecting the extent to which an individual cow is displaced by other cows during agonistic interactions), and the time cows spent standing half in the cubicles, predicted the subsequent likelihood of animals becoming clinically lame. Hopster et al. (1998) selected primiparous heifers with either high (“high responders”) or low (“low responders”) plasma cortisol concentrations in reaction to an OF test, and subjected these animals to an experimental intra-mammary endotoxin challenge (mimicking a mild mastitis infection) during the subsequent lactation (i.e., one year after the initial selection). High and low responders significantly differed in the numbers of circulating lymphocytes between 10 and 21 hours post challenge. In the context of the present thesis, this is an especially intriguing finding because it would suggest that underlying fearfulness (putatively reflected in the cortisol response to an OF test, see chapters 3, 4 and 6) may affect aspects of the immune response against mastitis in dairy cows. Interestingly, there was only a transient difference between high and low responders in plasma cortisol (one hour post challenge), supporting the assumption that other (e.g., neural or neuroendocrine) factors contributed to the difference in lymphocyte numbers between the treatment groups (Hopster et al., 1998). This, in turn, would concur with the notion of central (i.e., seated in the brain) rather than peripheral coordination of the response to an immunological challenge (such as exposure to endotoxin), mediated by underlying personality characteristics such as fearfulness. The present thesis provides some evidence that underlying sociality may also be relevant to dairy cow’s adaptation to regular husbandry conditions. The rate of OF and NO vocalisations at 7 months of age appeared to be correlated with inhibition of milk ejection during first machine milking more than one and a half years later (chapter 6). From a biological point of view, such a relationship would make sense on the assumption that vocalisation refers to a developmentally stable underlying trait that is capable of modifying the reactions of dairy cows to challenging situations at different ages and in different contexts.
Evaluating correlates of individual differences using a multidimensional approach
In the present thesis it is postulated that individual differences in stress responsiveness in dairy cows are mediated by multiple underlying traits (see above). A truly multidimensional approach in the assessment and appreciation of the implications of individual differences in stress responsiveness, however, would have required that relationships were examined between, on the one hand, measures reflecting the capacity of the individual cow to adapt to its living environment (e.g., with regard to health or inhibition of milk ejection), and, on the other, (co)variations in multiple traits obtained with the use of experimental challenges (i.e., the behavioural tests). In terms of the model in Figure 7.1, it would then be the position of an animal in a multidimensional space that determines its adaptive capacity, rather than its position relative to an individual dimension or trait (such as fearfulness, coping style or underlying sociality). This comes close to the concept of a multivariate “fitness landscape”, where “fitness” (e.g., survival, reproductive success, health, etc.) of animals in the wild is a function of multiple behaviours and their interaction (see Dingemanse and Réale, 2005, for an example with two behavioural traits “x” and “y”, measured on each individual, e.g., “aggressiveness” and “risk taking behaviour”, or “activity in the presence versus absence of predators”, etc.). In the current thesis, with only 23 heifers available for studying the relationship between early reactivity to behavioural tests and later responsiveness to a “real life” challenge (see chapter 6), the number of animals was deemed insufficient to allow for such a multivariate approach in a statistically reliable way. Other studies, however, have already indicated that this could be a useful approach. Cavigelli et al. (2009), for example, distinguished two independent traits in rats exposed to novel stimuli related to behavioural inhibition and glucocorticoid production, respectively, and demonstrated that the lifespan of rats that exhibited both high behavioural inhibition and high glucocorticoid production was almost two months shorter than that of rats that showed either high behavioural inhibition alone, or high glucocorticoid production alone. Likewise, in humans, the combination of “depression”, “hostility”, and “anxiety” was a stronger predictor of coronary heart disease than any of these personality traits alone (Boyle et al., 2006; Mehta and Gosling, 2008). Based on extensive work with lines of rats selectively bred for differences in their behavioural performance in an active avoidance test, Steimer et al. (1997) suggested that the propensity to develop mental disorders may rely on the interaction between two independent dimensions, i.e. “emotional reactivity” (ranging between high and low, i.e., similar to fearfulness) and “coping style” (either “active” or “passive”, and related to locomotor and rearing activity in a novel environment). Animals with a passive coping style and a high emotional reactivity (for a comparison, see lower right quadrant of Figure 7.1) would be susceptible to anxiety problems, whereas animals with an active coping style and a low emotional reactivity (see upper left quadrant of Figure 7.1) would be prone to impulsiveness and a lack of behavioural inhibition (Steimer et al., 1997).
Temperament, fitness and welfare
In the context of animal ecology, the term “fitness” generally encompasses two main aspects, i.e., reproductive success and survival rate, both of which ultimately determine the ability of the individual to propagate its genes (e.g., Barker, 2009). Differences in temperamental or “adaptive personality” traits, in turn, are believed to exert an important influence on “fitness” (see Sih et al., 2004a, b; Dingemanse and Réale, 2005; Réale et al., 2007; Dingemanse and Wolf, 2010; Wolf and Weissing, 2010). In the context of the breeding and husbandry of domestic animals, the term “fitness” is used rather generically, and may either refer to an equivalent notion of “robustness”, or to aspects like fertility, disease resistance, health, and longevity (e.g., Calus, 2006; Van der Werf, 2007; Gibbons, 2009; Goddard, 2009; Ten Napel et al., 2009). Regardless of terminology, the present discussion argues that, similar to wild animals, the “fitness” (or “robustness”) of domesticated farm animals, including dairy cattle, is mediated by stable underlying temperamental traits. In the case of dairy cattle, I suggest that at least three traits may be relevant in this respect: fearfulness, sociality and activity (or coping style). Because reduced “fitness” (or “robustness”) may be associated with problems that involve elements of mental suffering (e.g., pain in the case of disease), variations in these temperamental traits may also have implications for animal welfare (see Duncan, 2005; Broom, 2007; Dawkins, 2008).
Negative effects of selection for increased productivity
Although modern farm animals may not now be affected by most of the natural selection pressures that originally shaped the genotype of their wild ancestors, they are, however, continuously subjected to another major source of genetic change, resulting from selective breeding for production traits. Persistent artificial selection for production traits has tremendously increased production efficiency and production levels in all farm animal species over the last few decades (see Rauw et al., 1998; Sandøe et al., 1999; Oltenacu and Algers, 2005). At the same time, however, there is increasing evidence suggesting that intense selection for increased productivity has negatively affected the fitness of farm animals because of unfavourable genetic correlations between production traits and measures of health, fertility and longevity (see Rauw et al., 1998; Sandøe et al., 1999; Oltenacu and Algers, 2005; Oltenacu and Broom, 2010). The resource allocation theory is influential in explaining negative side-effects of selective breeding; this predicts that in farm animals, because of a disproportionate emphasis on biological processes related to production traits (e.g., growth or the production of milk), fewer resources (e.g., in terms of feed intake, body tissue, energy, etc.) are available for other important life functions such as reproduction or immune defence (Beilharz et al., 1993; Rauw, 2009). This may compromise the animal’s adaptive capacities, and render it susceptible to disease (Rauw et al., 1998; Mignon-Grasteau et al., 2005; Oltenacu and Algers, 2005; Jensen et al., 2008; Oltenacu and Broom, 2010).
Temperament and selective breeding for production traits
Comparative and genetic studies in poultry suggest that selective breeding for production traits may also affect temperamental traits such as fearfulness or sociality. For instance, in comparison with Red Junglefowl (the wild ancestor of domesticated layer breeds), modern White Leghorn laying hens behaved less actively in fear and exploration tests, and had a higher motivation to remain close to conspecifics (Schütz et al., 2001; Schütz and Jensen, 2001; Väsänen et al., 2005). Molecular genetic experiments, using progeny from an intercross between Red Junglefowl and White Leghorn lines, demonstrated that, to some extent, quantitative trait loci (QTLs) for such behavioural characteristics were located at the same genomic position as QTLs for production traits like growth and egg weight (Schütz et al., 2002, 2004), thus supporting a genetic basis for the link between selection for increased production and correlated behavioural changes. These findings were interpreted in terms of the resource allocation theory, and were thought to indicate that domesticated laying hens show less energy demanding behaviours than their wild ancestors, allowing them to reallocate the ‘saved’ energy to production traits (see Jensen, 2001, 2006, 2010; Jensen et al., 2008). Some studies support this idea. Selection for growth in beef cattle, for example, was correlated with slower flight speed (a behavioural measure of cattle temperament, see above), (Burrow and Prayaga, 2004). However, other reports imply an opposite phenomenon, i.e., more active and potentially energy demanding temperamental characteristics were linked with increasing production potential. In pigs, positive correlations were found between lean meat percentage and struggling behaviour during the backtest (Van Erp – van der Kooij et al., 2000, 2003), and between average daily gain and the level of aggression during a resident-intruder test (Cassady, 2007). In a more recent study, average daily gain in pigs was positively genetically correlated with the number of struggles during the backtest (Velie et al., 2009). Similarly, in some (but not in other) lines of laying hens, continuous selection for higher egg production was associated with increased and more intense behavioural and physiological responses to fearful stimuli, as well as higher levels of feather pecking and cannibalism (Van Hierden et al. 2002; Kjaer and Mench, 2003; Uitdehaag et al., 2008b). Apparently, reducing the expression of energy demanding behaviours is not the full story when it comes to explaining the effects of selection for higher production on temperamental traits. Van Hierden (2003) offered an interesting alternative hypothesis, and speculated that selection for higher productivity in laying hens may unintentionally target important neurobiological substrates of stress reactivity (e.g., the serotonergic system in the brain), thereby fundamentally changing the (central) neuroendocrine state and, hence, the behaviour and adaptive capacity of the animals. More recently, it was found that feather pecking in progeny from an intercross between Red Junglefowl and White Leghorn strains was linked to an active behavioural pattern in response to challenge (suggestive of a “proactive coping strategy”), as well as early sexual maturation, fast growth, weak bones, and, in males, a high fat accumulation, suggesting that feather peckers have a different resource allocation pattern (Jensen et al., 2005). Possibly, therefore, reallocation of resources and (unpredictable) changes of the neuroendocrine state under the influence of selection for increased production might both be results of this particular selection programme. This, however, remains to be demonstrated. Clearly, the foregoing examples illustrate that the effect of selective breeding on farm animal temperament and fitness is complex, and that, to date, underlying (genetic) mechanisms and pathways remain largely unknown (see Van der Werf, 2007).
Relationships between temperament, fitness and production
Figure 7.2 provides a schematic representation of (partly hypothetical) relationships between the main categories of traits that may affect the fitness of farm animals, including dairy cows. I will use this figure as a framework to further discuss the practical implications of the findings of the present thesis. Arrows (numbered 1 to 3) refer to (largely unknown) pathways and mechanisms underpinning each mutual relationship between categories of traits. Fitness is proposed to be equivalent to “robustness”, and to encompass a range of aspects including health, fertility, longevity and behaviour, e.g., normal, such as feeding, social or maternal behaviours, and abnormal, such as overt aggressive, harmful or stereotypic behaviours. Defined in this way, fitness determines the extent to which an animal successfully adapts to its environment, and, hence, its state of welfare (see also Moberg, 1987; Wiepkema and Koolhaas, 1993). It is assumed that temperamental traits may affect the fitness of farm animals through the interplay between components of the central nervous, neuroendocrine and immune systems (arrow numbered 1, see Figure 7.2). This was addressed in a previous section of this discussion (see above). Likely through a reallocation of resources (discussed above), selection for production traits may negatively affect the fitness of farm animals in more direct or proximate ways (arrow numbered 2, see Figure 7.2). For example, calcium and other minerals may be diverted from the process of bone formation to that of egg shell formation thereby affecting skeletal strength in laying hens, or bodily reserves may be mobilised to facilitate high milk production to the extent that dairy cows experience weight loss and a negative energy balance, etc. (see Rauw et al., 1998; Sandøe et al., 1999; Oltenacu and Algers, 2005, Veerkamp et al., 2009). It is also hypothesized that selection for production traits may affect temperamental characteristics (arrow numbered 3, see Figure 7.2), including fearfulness, sociality and coping style, because of a tendency to reduce the expression of energy demanding behaviours, and/or through effects on the neuroendocrine state of the animal, which may not necessarily be mediated by a reallocation of resources. This change in temperament, in turn, may again influence fitness (arrow numbered 1, see Figure 7.2). Thus, this “two-stage” route may represent more indirect or ultimate mechanisms by which selection for increased productivity may compromise fitness.
Balanced breeding and temperament
As a means of counteracting negative side effects of breeding for production traits on farm animal fitness and improving their welfare, adaptability and “robustness” it has been widely suggested that more “sustainable” or “balanced” breeding goals, including traits other than those strictly related to production characteristics in the selection index (i.e., multi-trait selection) should be defined (Sandøe et al., 1999; Kanis et al., 2004, 2005; Lawrence et al., 2004; Oltenacu and Algers, 2005; Calus, 2006; Star et al., 2008; Knap, 2009; Ten Napel et al., 2009; Oltenacu and Broom, 2010). Several scenarios have been explored in this respect. Many authors proposed using traits related to fitness (see Figure 7.2, upper right hand side), and, for example, suggested a selection index in dairy cattle based on traits like lameness, mastitis, calving interval and lifespan as measures of health and fertility (see Lawrence et al., 2004; Oltenacu and Algers, 2005). Traits of this kind have already been recorded in practice, and quantitative evaluations using real data show that, because of antagonistic relationships between production and fitness traits (arrow numbered 3, see Figure 7.2), a trade-off may exist between the costs of lower milk yield and the benefits of a higher health status of cows (Lawrence et al., 2004). Therefore, depending on the rate of genetic change, and the weights applied to each trait, breeding for improved welfare of dairy cows may be profitable overall (Lawrence et al., 2004). Other approaches have also suggested the use of behavioural or temperamental traits for breeding purposes, including, for example, fearfulness, sociality, or aggression (see Faure and Mills, 1998; Jones and Hocking, 1999; Kanis et al., 2004; Boissy et al., 2005a; Star et al., 2008). In terms of Figure 7.2, a breeding index may then consist of traits belonging to each of the three categories. In line with this idea, using temperamental traits in dairy cattle breeding may be one important potential way of implementing the results of the present thesis in practice. Until now however, in contrast to more “classical” fitness traits (see above), the application of temperamental characteristics in farm animal breeding is still largely a matter of theory, open for discussion.
Following the model in Figure 7.2, fitness would determine whether or not an animal is able to successfully adapt to its environment, and this raises the fundamental question of why temperamental traits should be considered in the first place. Ten Napel et al. (2009), for example, suggested focusing on “the results of adaptation” (i.e., fitness), rather than on “the adaptation process itself” when breeding for robustness in cattle. However, one reason to include underlying temperamental traits in a breeding programme would be that these traits are capable of influencing a wide range of biological responses to a broad variety of different challenges (discussed above), and, hence, capture more general and far-reaching features of the adaptive capacity of an animal than individual fitness traits. Moreover, temperamental traits are relatively stable across development, and may therefore predict the risk of a fitness problem before it actually occurs. In a breeding context, this could mean, for example, that selection based on a juvenile temperamental trait might target an adult fitness problem.
Figure 7.2 Schematic representation of interrelationships between the main categories of traits that determine the “fitness” (or “robustness”) of farm animals, including dairy cattle. Fitness encompasses a range of aspects, including health, fertility, behaviour (normal, e.g. feeding, social or maternal behaviours, and abnormal, e.g. overt aggressive, harmful or stereotypic behaviours), and longevity. Fitness defines the extent to which an animal successfully adapts to its environment. Arrows (numbered 1 to 3) refer to (largely unknown) pathways and mechanisms underpinning each mutual relationship. 1. Basic, underlying temperamental (or “adaptive personality”) traits, including fearfulness, sociality and coping style, are assumed to affect the fitness of farm animals through the interplay between components of the central nervous, neuroendocrine and immune systems. 2. Persistent selection for production traits, such as milk production or growth, may negatively affect the fitness of farm animals, for example, because of a shift in the allocation of resources. 3. An effect on temperamental traits may provide an additional mechanism by which selection for production traits may ultimately affect the fitness of farm animals. This effect may be the result of a reallocation of resources, and may involve the unintentional co-selection for specific neurobiological characteristics, thereby altering the neuroendocrine state of the animal. In addition to fitness traits, temperamental traits may represent potential selection crititeria to be included in breeding programmes aimed at improving fitness and welfare of farm animals, provided that the (genetic and phenotypic) relationships between production, temperament and fitness traits are elucidated and that the underlying biological mechanisms are better understood.
Appropriate temperament A second important question that emerges when evaluating the practical feasibility of incorporating temperament traits in farm animal breeding concerns the assessment of the desired or appropriate kind of temperament. Several authors seem to assume a priori that selection for temperament should aim at reduced fearfulness or increased sociality (e.g., Faure and Mills, 1998; Jones and Hocking, 1999; Boissy et al., 2005a). Kanis et al. (2004) addressed this issue in a more conceptual way, and described a framework for breeding for improved welfare in pigs, using a thermoregulatory model as a basis. The width of the thermoneutral zone, i.e., the range of ambient temperatures in which little or no behavioural or physiological effort is required to maintain a constant body temperature, was assumed to be positively associated with animal resilience and welfare (Kanis et al., 2004). This principle was extrapolated to other environmental stimuli such as, for example, novel, social, or otherwise (emotionally) demanding situations, and led to the prediction that pigs that exhibit a low average response to a range of challenges (i.e., low aggression, low exploration, low fear, low “nervousness”, etc.) experience high resilience and welfare and vice versa (Kanis et al., 2004). From an animal welfare point of view, a desired pig temperament would then by definition refer to animals with low amplitudes of behavioural and physiological responses to challenging situations (i.e., situated on the left-hand side of the distribution presented in Figure 7.1). These assumptions, however, may fall short in a number of ways: Qualitative dimensions. First, they overlook the fact that behavioural and physiological response patterns of (farm) animals to challenge may also include qualitative dimensions (i.e., the type of response, or coping style, at a given amplitude), that may be important with regard to fitness and welfare (see previous sections of this discussion). Amplitude of response and adaptation. Second, it is questionable whether low amplitudes of response to stressors are always favourable from an adaptation or fitness point of view. For instance, according to recent theory on robustness of biological systems, the robustness of an organism does not mean that it remains unchanged in response to environmental stimuli; on the contrary, a robust organism is able to mount an adequate and flexible response to a changing environment (Kitano, 2004). This closely agrees with the concept of allostasis, which emphasizes the notions of “maintaining stability through change”, and providing the animal with the appropriate biological make up (e.g., behaviourally, physiologically, and neuroendocrinologically) to be able to do so (see McEwen and Wingfield, 2003; Korte et al., 2007; Feder et al., 2009). Recent examples of putative (genetic) relations between temperament and fitness (or “robustness”) also support this view, and point to the possibility of a link between a high amplitude and intensity of response to challenge and high fitness. Canario et al. (2009) showed that pigs with a beneficial genetic effect on the growth of pen-mates (using a novel approach to quantify the heritable effects that animals have on their group mates’ traits, see Rodenburg et al., 2010, for a review) spent more time fighting and bullying others at mixing, initiated more fights, and also both won and lost more fights. It was suggested that the behaviour of these pigs may benefit pen-mates by speeding the establishment of dominance relationships (Canario et al., 2009). Likewise, Gibbons (2009) compared the behaviour during feeding of daughters (primiparous heifers) of bulls with either a high (“high index”) or low (“low index”) value of a “robustness index” which was an extension of an existing “Profitable Life Index” in the UK, with increased emphasis on locomotion, somatic cell count, udder health, fertility and lifespan. High index daughters were found to respond more frequently to aggressive interactions (i.e., either by active avoidance or by retaliation) than low index ones (Gibbons, 2009). The present results suggest that a high level of sociality, putatively reflected in a high rate of vocalisations during OF and NOT tests (which equates with a high amplitude of response), may be beneficial to milk ejection at the beginning of lactation in heifers (see chapter 6). Conversely, in other cases low amplitudes of response might be more adaptive than high ones. For example, low aggression and high docility of Limousin heifers during standard encounters with a human handler were genetically associated with good maternal behaviour and high fertility (Phocas et al., 2006).
Genotype x environment interaction. Third, the theory of “adaptive personality differences” or “behavioural syndromes” would predict that the fitness consequences of differences in temperament directly depend on the environmental conditions the animals are exposed to (see Sih et al., 2004a, b; Groothuis and Carere, 2005; Wolf and Weissing, 2010). In other words, from a fitness point of view, a certain response pattern resulting from a particular underlying temperament (involving a specific type or amplitude of response to challenge) could be (highly) beneficial in one environment but less beneficial or even detrimental in another (see also previous sections of this discussion). In population genetic terms, this is described as a “genotype x environment (G x E) interaction” (see Calus, 2006). A number of studies clearly suggest that G x E interactions are also relevant for farm animals, and, hence, for the effects temperamental traits may have on fitness (arrow numbered 1, see Figure 7.2). Bolhuis et al. (2003, 2006), for example, demonstrated that differences in immune response to antigenic challenge, or in the prevalence of gastric lesions recorded at slaughter, between pigs previously characterised in a backtest as either high or low responders, depended on the housing system (barren versus enriched) the animals were kept in. These findings provide examples of significant temperament x environment interactions at the phenotypic level. In quantitative genetic work in dairy cattle, significant G x E interactions were reported for several putative fitness traits, including body condition score, the number of inseminations before conception, and survival (reflecting whether or not a cow is present on the farm during the next lactation) (Calus et al., 2005).
Collectively, therefore, it is suggested that an appropriate temperament is all about a response to challenge appropriate for the environmental conditions, rather than a certain magnitude or qualitative type of response per se. Or, in terms of the model in Figure 7.1, various response patterns situated at different positions in the multidimensional space (or “fitness landscape”, see discussion above) may all be associated with an acceptable level of fitness (and welfare), given the right circumstances. At this time, we are still a long way from understanding the biological mechanisms underpinning the relationships between temperament, fitness and selection for enhanced productivity in dairy cattle. Moreover, in contrast to aspects of dairy cow fitness like fertility, health and longevity, temperamental traits, in the sense of the present thesis, have not been recorded in any significant numbers of cows so far. This precludes an immediate application of the current findings in practical dairy cattle breeding.
Practical way forward and prospects
What is needed are studies where the three categories of traits indicated in Figure 7.2 are simultaneously recorded in large numbers of animals in practice, preferably in different environments (e.g., conventional versus organic farms, see Nauta et al., 2006). This will enable quantitative genetic analyses and the estimation of genetic parameters, including phenotypic and genetic correlations between production, temperament and fitness traits. With this information, accurate simulation studies could conceivably identify feasible strategies on the utilization of temperament traits for the improvement of robustness and welfare of dairy cows (see Kanis et al., 2005, and Gourdine et al., 2010, for examples in pigs). However, since the execution of behavioural tests according to the experimental approach described in this thesis is both time- and labour consuming, the practical feasibility of the required large-scale studies may be limited. Thus, there is a parallel need for the development of simpler and easier methods of recording relevant temperamental traits in dairy cattle. Perhaps the application of sensor-based methods may be helpful in this respect. Schrader (2002), for example, found significant relationships between temperamental characteristics observed during behavioural tests, and measures of dairy cow activity recorded with the use of pedometers (activity monitors) attached to the hind leg (see also Müller and Schrader, 2003). It is tempting to speculate that with the use of sophisticated sensors for monitoring behavioural (e.g., activity) and physiological measures (e.g., temperature, heart rate), aspects of dairy cow temperament might ultimately be tapped on a routine basis (see Berckmans and Guarino, 2008).
Encouragingly, the available data seems to indicate that heritabilities of temperamental characteristics in farm animals, including grazing species such as (dairy) cattle and sheep, are reasonably high, and mostly well within the range reported for production traits. For example, heritabilities of the level of docility of Limousin heifers during standard encounters with a human handler (Le Neindre et al., 1995), the movement of cattle when individually confined on a weighing platform (Schmutz et al., 2001), and vocalization and locomotor responses of sheep to social isolation or fear of human tests (Boissy et al., 2005b; Wolf et al., 2008), all ranged between 0.22 and 0.58. Similar heritabilities for various behavioural (e.g., OF locomotion and vocalization, latency to first contact with a novel object and the time spent in contact with it) and physiological (e.g., heart rate and heart rate variability during confinement in a start box, and the cortisol response to an OF and NO test) measures were obtained in a recent study in dairy calves, using exactly the same protocol for OF and NO testing as described in the present thesis (see chapter 4) (Van Reenen et al., 2008; Eaglen, 2009). Notably, there are also indications that the phenotypic relationship in calves between avoidance of a novel object and the cortisol response to OF and NO tests, (as reported in this thesis (see chapters 3 and 6), also exists at the genetic level (Eaglen, 2009). These findings support the idea that correlated adrenocorticortical and behavioural responses to novelty, and OF locomotion and vocalization reflect genetically mediated temperamental traits in dairy cattle. This merits further study, and may ultimately justify large scale (genetic) research.
The present thesis reports consistent individual differences in behavioural and physiological responses of dairy cattle to acute experimental stressors, both during the rearing period (3 weeks to 6-7 months of age), and from the rearing period to (early) adulthood (22 - 29 months of age). In addition, it was demonstrated that dairy heifers exhibited consistent differences in their response to a “real life” challenge such as being machine-milked for the first time, and that part of this variation was explained by differences in earlier behavioural responses to OF and NO tests. These findings significantly strengthen the idea that stress responsiveness in dairy cows is controlled by developmentally stable underlying temperamental characteristics. The multidimensional nature of response patterns to challenges observed in dairy cattle in the current work is supported by reports in other species. Results of the pharmacological validation study, using the fear-reducing agent brotizolam, confirmed the hypothesis that temperament in dairy cows consists of multiple independent dimensions, and identified underlying fearfulness as one of the candidate traits. Other traits likely include sociality and activity (or coping style). Like other (farm) animal species, traits of this kind may affect the fitness of dairy cows in terms of, for example, health, fertility and longevity. In addition to “classical” fitness traits in dairy cows, such as calving interval, lifespan, or measures of lameness and mastitis, temperamental traits may represent additional selection criteria that should be included in breeding programmes aimed at the improvement of “robustness” and welfare. This will require the elucidation of (genetic and phenotypic) relationships between production, temperament and fitness traits, as well as further insight into the biological mechanisms underpinning these relationships. Such knowledge might enable the breeding of dairy cows that are optimally adapted to their environment, including alternative husbandry systems that are intended to improve animal welfare and sustainability (e.g., organic farming systems). Prior to the actual implementation of temperamental traits in breeding programmes, selected behavioural and/or physiological measures of dairy cow temperament might be immediately useful as tools to monitor the potential consequences of selective breeding for adaptive capacity and welfare.